Hygraph

Hygraph offers 15+ field types including JSON, location, color, enumerations, union (polymorphic) relations, components and remote fields, and the Aug 2026 Schema Import copies models, components, remote sources, enumerations and taxonomies between projects. Schema can be scripted through the Management SDK migrations, but there is no declarative schema-as-code source of truth, which keeps it below the top tier.

Relations are graph-native: two-way references are generated automatically and queryable from both ends, many-to-many is native, and union relations allow several models per field. Content Federation adds remote fields that resolve live external data inside the same query. Only lacking relation-level metadata keeps it short of the very top.

Components and modular (multi-type) component fields allow nested block composition, and Rich Text can embed entries and assets with AST output. Live Preview now tracks fields inside nested components (Jul 2026), and Variants reuse the same structure. Nesting depth and component limits are plan-capped (10/20/up to 150 components), so it sits just below Portable Text class leaders.

Built-in validations cover required, unique, regex, character and range limits and list counts, enforced through the Content API as well as the editor; Jul 2026 brought inline validation for multi-value fields and clearer required-field errors. The new relation and union editorial rules are explicitly not enforced by the API, and there is no cross-field or custom rule engine beyond app extensions.

Versions are created on every publish, with a side-by-side compare (show-only-edits) and field-by-field restore. Draft content is not versioned, versioning exists only on paid plans (14 days on Growth, up to 365 on Enterprise), and versions do not store referenced entries or assets, so deleted content cannot be restored from history.

Taxonomies are hierarchical node trees attached to models through a dedicated Taxonomy field (single or multi-value), reusable across models, queryable through the API and importable between projects. They are Enterprise-only (up to 12 per project), and synonyms or term-governance workflows are not documented, so it lands just under the 75 band.

Live Preview shows draft content on the customer's own frontend with optional real-time field updates, and Click to Edit adds hover overlays and Edit buttons via the preview SDK with guides for Next.js (App and Pages Router), Remix, Vue/Nuxt and vanilla JS, including variants. Click to Edit is still in beta and opens the field in Studio rather than editing inline on the page, and it needs data-hygraph-* attributes wired by developers.

Rich Text returns a portable AST alongside HTML, Markdown and text, supports embedded entries and assets, tables and custom element rendering through the official React and HTML renderers. Custom marks and annotations are limited compared with Portable Text editors, which keeps it at the top of the adequate band rather than best in class.

The rebuilt Asset Management System offers URL-based transformations (resize, crop, WebP/AVIF, quality), localized assets, GraphQL upload and an improved asset picker. Aug 2026 added safeguards that block transformations on sources above 100 MP and rate-limit first-time transforms, and folder organization and usage tracking remain thin.

Collaborator View shows the avatars of everyone viewing the same entry, and commenting supports assignment, resolve and @mentions with notifications on all plans. There is no simultaneous co-editing; concurrent edits are handled by save conflict detection rather than live merging.

Custom Workflows (up to 10) route entries through steps with role-based assignees, can include AI agent steps, and work with up to 5 content stages and stage-restricted publish permissions; Content Freeze blocks changes in critical windows and audit logs record actions. All of this is Enterprise-only (Hobby and Growth get only commenting and assignment), and conditional routing is absent.

Scheduled publishing and unpublishing work for single entries or bundled releases, with a Releases overview listing scheduled items (no longer capped at 10 since Jul 2026) and Content Freeze for embargo windows. The feature is Enterprise-only and there is no true publishing calendar view, so it misses the 75 band.

Localization is field-level by design, with per-locale staging and publishing, localized assets, fallback chains passed as a locale array (or the gcms-locales header), and locale-scoped permissions. Locale counts are tight below Enterprise (2 on Hobby, 3 on Growth, up to 80 on Enterprise, $150/month per extra locale).

Custom roles (Enterprise, up to 30) scope content permissions by model, action, locale, stage and conditional entry filters, with environment-specific content permissions and new granular management permissions in 2026; SAML SSO via Okta, Entra ID and PingFederate is Enterprise-only. Field-level permissions and SCIM provisioning are not documented, and Hobby/Growth get only fixed system roles.

GraphQL-native Content API with separate Management API, rich filtering, ordering, pagination, locale and stage arguments, and Content Federation joining remote REST and GraphQL sources in one query. The high-performance endpoint adds partial cache invalidation. The lack of a REST delivery fallback costs the usual few points.

Content and assets are served from a global CDN with automatic invalidation on mutation and partial cache invalidation by model and stage on the high-performance endpoint, plus regional clusters and dedicated infrastructure on Enterprise. There is no customer edge compute or edge-side personalization.

Webhooks cover create, update, publish, unpublish and delete with filtering by model, stage, action and source, HMAC-SHA256 signed payloads, custom headers, up to 5 retries and delivery logs. There is no native event streaming target (EventBridge, Pub/Sub) and schema-change events are limited.

Hygraph is purpose-built headless: channel-agnostic GraphQL, AST rich text with multiple renderers, federation of external sources, and starters across web and mobile frameworks. Official SDKs are concentrated on JavaScript/TypeScript, relying on standard GraphQL clients elsewhere, so it does not reach the top of the band.

Hygraph Segments are named labels only; the docs state the actual user lists are defined and managed in an external system, and Segments are Enterprise-only (up to 500). Audience data comes through Dynamic Yield, Optimizely or Segment marketplace apps, with no profiles or segment builder in the product.

Variants (up to 30 per entry, Enterprise) let editors author per-segment versions of an entry in Studio, and Click to Edit works on variants in preview. Choosing which variant a visitor sees is left to the frontend or an external decision engine, which keeps it in the middle of the adequate band.

Experiments run in Optimizely or Dynamic Yield through marketplace apps, with Variants supplying the content. There is no traffic allocation, statistics or results reporting in Hygraph itself.

No recommendation engine is documented; related content must be curated manually through relations or taxonomies.

Visitor-facing search relies on Algolia, Elastic or Klevu marketplace apps and webhook-driven indexing; the Content Finder and _search filters are editor or API features, not a site search product. Connectors exist but index sync and UI components are thinner than a first-party search integration.

Hygraph has no catalog, cart, pricing or checkout; commerce is positioned as an external source federated into the content graph.

Remote Sources federate live product data from commercetools, Shopify, BigCommerce, SAP and Salesforce Commerce into GraphQL queries, backed by product-picker marketplace apps. That depth would score in the 60s, but it is an integration path and is held at the ceiling; Remote Sources are limited to 0 / 1 / up to 10 by plan.

Built-in analytics are operational: a usage dashboard with API and safeguard charts and project analytics in the new project directory. No content performance or engagement analytics exist.

A Google Analytics marketplace app and a Segment app exist, plus Zapier and n8n connectors for event flows, but tagging and event tracking still live in the customer's frontend. Coverage of Adobe Analytics or Amplitude is not documented.

Multiple sites are usually modelled as separate projects or environments, now helped by Schema Import (Aug 2026) and copying entries between projects (Sep 2026); multitenancy is listed on Enterprise. There is no native shared-content model across sites, so it reads as silo-based multi-site.

Marketplace apps connect Smartling, Lokalise, Crowdin, DeepL and EasyTranslate, and Studio added copy-to-all-locales (Jul 2026). Translation workflows run through those connectors rather than an in-platform translation job manager, keeping it below the integration ceiling.

Organization-level management, per-project custom roles and Schema Import help keep brands consistent, but there is no shared component library with brand overrides, cross-brand approvals or global policy enforcement.

The native Asset Management System is a solid library with localized assets and transforms but lacks rights, expiry, approvals and brand portals. Enterprise DAM comes through Hygraph-built Cloudinary and Bynder apps, which are integrations capped at 60.

Video hosting, transcoding and streaming come through the Mux app with in-editor upload; Hygraph itself stores video files without transcoding or adaptive delivery.

Pages are assembled from modular component fields in the entry form, with components reorderable by drag (Jul 2026) and Live Preview showing the result. There is no drag-and-drop canvas, template locking or visual composition on the frontend, so it is a structured block editor rather than a page builder.

Hygraph is headless with no rendering layer; frontends are hosted on Vercel or Netlify using official marketplace apps, deploy guides and starters. The vendor does not host the delivered site.

There is no form builder; the documented approach is creating submissions programmatically through mutations from a custom frontend.

Salesforce Marketing Cloud and Klaviyo apps allow content reuse in email tools, and other ESPs are reached through webhooks or Zapier. No subscriber sync, triggered sends or email preview exist in Hygraph.

No marketing automation exists; webhooks to an external automation tool are the only route.

SEO is modelled as fields in the schema, with slug generation from templates (now including related-model fields, Aug 2026). Sitemaps, redirects, canonicals and structured data are built by the frontend team; the SEO agent is an AI feature scored elsewhere.

The product offers nothing for answer-engine readiness of published sites: no llms.txt generation, Markdown endpoints or crawler controls, since Hygraph does not serve the site. The SEO agent's GEO review reports (Jul 2026) are an AI authoring feature, not a visibility capability of delivered content.

The marketplace lists 44 apps, many built by Hygraph, spanning commerce, DAM, translation, search, SSO and hosting, with an App Framework for custom apps. It is well curated but small next to marketplaces with 100+ integrations.

Hygraph has no visitor identity concept; the SSO apps (Okta, Entra ID, PingFederate) cover CMS users only. Gated content means the frontend authenticates visitors and filters content with scoped tokens.

Hygraph's GraphQL-native API remains the platform's signature strength: schema is auto-generated from the content model, fully introspectable/self-documenting, and supports nested selection, filtering, ordering, cursor and offset pagination, and relationship traversal in one request. Content Federation extends the schema with remote sources for a unified content-plus-external API surface, and the 2026 high-performance Content API endpoint adds GraphQL Mutations alongside the REST Management API (redesigned 2026 for stability). The only material gap is the absence of a REST content-delivery option for REST-only toolchains.

The 2026 high-performance Content API endpoint adds model-and-stage-based partial cache invalidation, so content updates purge only affected cache segments, on top of global CDN delivery and cursor pagination for large datasets. Uncached request limits are now published per plan (5 RPS Hobby, 25 RPS Growth, up to 500 RPS Enterprise), and August 2026 added a per-environment RPS limit on first-time image transformations plus a 100-megapixel transformation cap. The low uncached ceilings on self-serve tiers and query-design-dependent latency for deep GraphQL queries keep it out of the mid-80s.

JavaScript/TypeScript remains the primary SDK surface, with the official Management SDK (Jan 2026) adding TypeScript-first programmatic project/schema control on top of existing GraphQL clients (graphql-request, Apollo, urql) and the Preview SDK. Rich Text renderers exist for React and HTML. There are still no official SDKs for Python, Ruby, Java, .NET, PHP, Swift, or Android — teams use standard GraphQL clients in their preferred language, which works but limits scaffolding and language-specific community resources.

The marketplace has expanded through 2025-2026 with new connectors spanning translation (Lokalise, Crowdin), DAM (Bynder, Cloudinary), commerce/PIM (Inriver), analytics (Segment), automation (Zapier), CRM/marketing (Salesforce Marketing Cloud), and visual building (Plasmic) — now covering most major categories, though still well short of Contentful's 100+ catalog. Content Federation partially offsets the gap by enabling direct API federation rather than requiring a marketplace connector. Depth and count keep it in the mid-range rather than the 75+ band.

Extensibility centers on the App Framework (App SDK), successor to the deprecated UI Extensions, for custom field editors, sidebar widgets and page extensions, installable from a local app during development. Webhooks, Content Federation for server-side schema extension, the Management SDK for config-as-code, and a remote MCP server (Claude connector added August 2026, with agent runs exposed) round it out. There are still no in-platform server-side hooks or custom API endpoints inside Hygraph, holding it below 75.

SSO via SAML 2.0 and OIDC available on Enterprise plans. MFA (two-factor) available for all users. Public Content API uses Permanent Auth Tokens (PATs) with optional content permission scoping; Management API uses PATs or service account tokens; OAuth-based access for App Framework apps. SSO being Enterprise-only is a mid-market friction point, consistent with most headless CMS peers.

Custom roles with per-model, stage, locale and environment CRUD and publish permissions, field-based CRUD permissions (2026), and scoped Permanent Auth Tokens give fine-grained control. July-August 2026 added more granular management permissions (Can see project settings, Can see environment switcher, Can see apps section), permission-gated Content Freeze, and read-only scoped MCP tokens. Content-instance access control (entries visible only to their owner or brand) is still not native, holding it below 80.

Hygraph holds SOC 2 Type 2 (re-audited Jan 2026), ISO 27001-certified infrastructure, and GDPR compliance with a DPA (including SCCs and published sub-processor list) available to all customers, plus EU data residency from its Berlin base. No HIPAA BAA and no FedRAMP authorization, so healthcare and US federal buyers require careful evaluation. The SOC 2 Type 2 + ISO 27001 + GDPR + EU residency combination hits the 80+ band for headless CMS peers.

No major publicized security breaches in Hygraph's history. The company maintains a responsible disclosure policy (security@hygraph.com) and is AWS-hosted, but there is no public bug bounty program. As a smaller company than Contentful or Salesforce, security communications are less comprehensive than larger vendors, and limited public information keeps confidence LOW.

SaaS-only with no self-hosted or customer-managed private cloud option. Region choice is broad for a SaaS CMS: 3 shared regions (EU, US, APAC) on self-serve plans, 9 shared-cluster regions and 34 dedicated-cluster regions on Enterprise, with read replicas as an add-on. That eases data residency but does not help buyers who require on-prem or self-managed deployment.

Enterprise contracts carry an uptime SLA of up to 99.95%, backed by a public status page (status.hygraph.com) and CDN redundancy for cached reads; Free and Growth plans have no SLA. After acknowledged outages on April 30 and June 26, 2026, the July to September quarter was quiet: the only posted incident was a 35-minute sign-up outage on August 7 that did not affect existing projects. Plan-gated SLA and the earlier 2026 incidents keep it below 80.

Hygraph serves enterprise customers at scale with CDN-backed delivery handling high read throughput on AWS infrastructure, and the 2026 partial cache invalidation improves cache efficiency under high update volume. Auto-scaling is managed by Hygraph; customers have no independent horizontal scaling controls, and GraphQL query complexity becomes a performance consideration at high scale requiring efficient query design. Entry and API call limits are generous on paid plans.

Automated backups with one-click recovery (Enterprise), multi-AZ infrastructure, and full content, asset and schema export through the Management API and Management SDK. Exports use Hygraph's own format and need transformation for other platforms. RTO/RPO figures are not published and one-click recovery is Enterprise-gated, holding it below 75.

There is still no local Hygraph server or emulator — development runs against a remote project (typically a dev environment) via the Hygraph CLI — but App Framework apps can now be installed and tested directly from a locally running application, improving the extension development loop. Because Hygraph is GraphQL-native, any GraphQL IDE (GraphiQL, Altair, Insomnia) works locally against the remote endpoint, and introspection enables offline schema tooling (type generation, validation). The structural lack of a content emulator keeps it mid-range, matching Contentful/Sanity peers.

The Management SDK enables TypeScript-first schema migrations, including atomic batch migrations, and config-as-code; Environments isolate dev, staging and production; webhooks trigger downstream pipelines. August 2026 Schema Import copies models, components, remote sources, enumerations and taxonomies from another project or template into a target environment without touching existing schema, and June 2026 Content Freeze can lock content during release windows. Environment promotion still requires scripting rather than one-click promote or aliasing.

Documentation is well-organized with framework-specific getting-started guides (Next.js, Nuxt, Astro, SvelteKit and others), an embedded GraphQL playground in the console for interactive query exploration, and an API reference auto-generated from the GraphQL schema. The 2026 Examples library expanded code samples for content delivery, mutations, and secure server-side patterns. Some advanced areas (Content Federation patterns, complex modular content) still warrant deeper coverage.

First-class TypeScript support across the stack: the GraphQL schema generates TypeScript types via graphql-codegen for end-to-end type safety from content model to frontend (including typed operations and Apollo hooks); the Management SDK is TypeScript-first for programmatic schema/config code; and the Preview SDK ships typed APIs. This gives Hygraph one of the stronger TypeScript stories among headless CMS platforms — auto-generated content types plus typed administrative tooling.

Cadence remains high and steady: the structured changelog shows a dated release every month from April through September 2026, with recent drops including Scheduled Publishing, Schema Import across projects, cross-project entry copying, the Hygraph MCP connector in Claude, a GEO review mode for the SEO agent, and new role permissions, on top of the 2025 to early 2026 re-platform, AI Agents and AI Assist launches. Not higher because releases are batched monthly rather than shipped continuously with per-fix notes.

The changelog is now a dated monthly log split into New features, Improvements and Bug fixes, with links to the relevant docs (e.g. Import schema, Transformation safeguards) and plain notes on behavioral changes such as the new 100-megapixel transformation cap and per-environment rate limit. That closes the earlier gap of minor fixes going undocumented. Not higher because there is no dedicated breaking-change label or per-release migration callout.

Hygraph communicates roadmap direction via blog posts, conference talks, and enterprise briefings, and named upcoming 2026 items publicly (flexible field type, customizable asset model with custom metadata) alongside early-access programs. There is no fully public roadmap with community voting (unlike Sanity's open approach). Not higher because multi-quarter planning visibility still typically requires Enterprise contact.

GraphQL API stability is high because the delivered API surface is generated from user-defined schemas, so Hygraph rarely breaks its own contract; the 2026 platform re-architecture shipped with a published migration path rather than a hard cutover. SDK updates follow semver and the Management API is versioned with backward-compatibility guarantees. Not higher because there is no formal codemod tooling or publicly documented 12+ month deprecation-window policy.

Hygraph's community is active but materially smaller than Contentful's or Sanity's. A community Slack is the primary public channel; the official Management SDK pulls roughly 8.8K npm downloads a week, and G2 review volume (~725) is strong relative to platform scale, but Stack Overflow coverage for Hygraph-specific tags is thin. Not higher because community footprint trails Tier 1 headless CMS leaders.

Hygraph DevRel is responsive in Discord with relatively quick answers, and GitHub issues on SDK repositories receive prompt maintainer attention. Conference presence at JAMstack Conf and GraphQL Summit reflects genuine engagement. Not higher because engagement quality, while strong, is bounded by the smaller community size — answer volume and depth of community-contributed content trail Tier 1 peers.

Hygraph's Solutions Partner program expanded meaningfully in 2026 with a wave of named SI/agency partners (Innowise, Turbine Kreuzberg, Macaw, codecentric AG, Shopmacher, Emerge Interactive, and others) plus a formalized program offering 10% revenue share, priority training, and demand generation. Technology partnerships span Vercel, Netlify, and major clouds. Not higher because the network still skews to DACH/EMEA boutiques and lacks the certified breadth of Contentful or a marquee global SI (Accenture/Deloitte) at scale.

Tutorial coverage spans official docs, the Hygraph YouTube channel (~10K subscribers), and community blogs covering Next.js, Nuxt, Astro, and SvelteKit integrations; transferable GraphQL content extends the effective surface. Not higher because Udemy/Pluralsight coverage is minimal and some legacy content still references the old GraphCMS brand, creating discoverability friction.

Finding developers with Hygraph-specific experience is harder than for Contentful, but the underlying skillset (GraphQL, TypeScript, modern JS frameworks) is broadly available, and a competent GraphQL developer can become productive in days. Not higher because the named-platform skill is still niche — LinkedIn endorsements and Hygraph-tagged job postings remain a small fraction of the headless CMS market.

Strong, sustained momentum: ~725 G2 reviews, 65 badges across 97 G2 Winter 2025 reports (published Jan 2026) including #1 Enterprise Usability, #1 Implementation for Headless CMS (five consecutive reports), and #1 User Satisfaction, plus a visible 2026 partner-signing streak and heavy AI-launch press. Not higher because enterprise logo announcements and case-study cadence still trail Contentful and Sanity.

Hygraph (formerly GraphCMS) has raised ~$43.7M total (Series B $30M led by One Peak, 2023) and shows no distress signals or layoffs through mid-2026, with a new CEO (Will Robinson) appointed Jan 2026 to lead an AI-focused next chapter. Not higher because there has been no fresh growth round since 2023, headcount sits modestly (~65–95), and a CEO transition without new capital is a watch-item rather than a scale-up signal.

Hygraph holds a clear, defensible position as the GraphQL-native, federated content platform, now reinforced by an AI-agents narrative, and sits in the G2 Leader quadrant. Differentiation against Contentful, Sanity, and Storyblok is sharp. Not higher because larger competitors are adding GraphQL surfaces and Hygraph still lacks a top-right Gartner MQ or Forrester Wave placement to anchor enterprise short-lists.

Sentiment remains strong: G2 4.4/5 across ~725 reviews with #1 User Satisfaction and five consecutive #1 Implementation awards, praising GraphQL API quality, content-modeling flexibility, fast onboarding, and support responsiveness. Not higher (and nudged from 78) because the aggregate rating eased to 4.4, and recurring criticisms — pricing (free→Growth jump), a steeper non-technical learning curve, no REST option, and smaller ecosystem — appear consistently in negative reviews.

Hobby ($0) and Growth (from $199/month, monthly or yearly billing) are fully priced with a detailed limits table, overage rates are stated, and the pricing FAQ now publishes an Enterprise range of about $32,000 to $300,000 per year plus a $150/month per-locale add-on price. Publishing an Enterprise band is rarer than the sales-gated norm, which lifts this above the 55–70 range. Not higher because the actual Enterprise quote, AI Agents add-on (ballparked at ~$10,000/year) and AI Assist overages are still finalized case by case.

Growth is metered on API operations and asset traffic, with overages billed automatically at the end of the cycle at USD 0.20 per 10,000 API operations and per GB of asset traffic, which is USD 20 per million calls; the earlier figure of $0.20 per million understated this a hundredfold. A Growth plan includes only 1M API operations, so a build-heavy Next.js site or a traffic spike can add a bill comparable to the base fee, and there is no tier between $0 and $199. Not lower because cached requests are not rate-limited or blocked, Hobby hard-stops at its limits rather than charging, and Enterprise is built from a platform fee plus negotiated usage dials.

The current comparison table gates far more than SSO: custom roles (0 on Growth), scheduled publishing, custom workflows, taxonomies, variants, automatic backups with one-click recovery, audit logs, an uptime SLA and an IT security review are all Enterprise-only. Remote Sources is not available on Hobby (0) and Growth gets just one, so the federation feature is effectively an Enterprise capability, contrary to the earlier finding. Not lower because core modeling, live preview, commenting, webhooks, environments and the full API are available on every tier.

Growth is self-serve on monthly or yearly billing with a credit card, so teams can start and stop without a contract, and Hobby projects can switch to a 30-day Enterprise trial without a card. Special pricing for students, nonprofits and open-source projects is offered on request, and Enterprise locales can be added by contract amendment without renegotiating. Not higher because Enterprise is an annual agreement with no published exit terms, and the discount programs are case-by-case rather than published.

Hobby is free forever with no stated commercial-use restriction: 1,000 entries, 500,000 API calls, 100 GB asset traffic, unlimited asset storage, 3 seats, 2 locales, 20 models and live preview. Usage is blocked rather than billed at the limit, so the tier cannot generate surprise charges. Not higher because assets count against the 1,000-entry cap, there is no version history, Remote Sources is unavailable on Hobby, and the 5 req/s uncached rate limit forces throttling during parallel builds.

The GraphQL API is auto-generated from the content model, so a first content query can run within minutes of defining a schema, and the in-console GraphQL Playground enables immediate API exploration with no local setup. G2 has ranked Hygraph #1 in Implementation for headless CMS multiple times, directly validating fast onboarding. Not higher because developers new to GraphQL face a modest query/schema learning curve before productivity.

Community-reported timelines are competitive: 3–6 weeks for a marketing site, 6–12 weeks with federation and multi-locale, 2–4 months for enterprise builds. GraphQL schema auto-generation accelerates the data-layer phase, while Content Federation configuration adds effort for commerce-integrated projects. Held at 70 because faster-than-average setup (G2 Implementation awards) offsets but does not eliminate the GraphQL learning curve.

The core skill required is GraphQL (querying, schema design, mutations) plus React/Next.js and TypeScript — all mainstream, so a senior frontend developer becomes productive within about a week. Hygraph-specific knowledge (Management API, Content Federation, UI Extensions) adds a modest premium over generalist rates. Not higher because GraphQL specialization still narrows the pool slightly versus pure REST/Contentful developers; not lower because GraphQL is now widely adopted.

Fully managed SaaS with no servers, databases, or CDN to provision — the subscription is the primary infrastructure cost, and asset storage/delivery are bundled within plan limits. Frontend hosting (Vercel, Netlify, Cloudflare Pages) is separate but platform-agnostic and standard for any headless setup. Scores high on the standard SaaS advantage of bundled, predictable infrastructure cost.

The managed SaaS model puts patching, scaling, and availability on Hygraph, so a single developer can cover operations part-time with no dedicated DevOps or DBA. The main ongoing tasks are monitoring API usage to avoid overages and managing schema changes across environments — the rate-limit throttling some teams add for parallel builds is the only notable friction. Not higher because usage monitoring is a real, recurring operational responsibility.

Content, assets, and schema are exportable via the Management API in Hygraph's JSON format, so data is accessible without a support ticket. However, Content Federation configurations, Rich Text stored as a proprietary AST, and UI Extensions all require transformation or redesign on a target platform, and the GraphQL-native delivery model has few equivalent destinations. Exit cost is moderate — data is retrievable but a real migration takes meaningful engineering effort.

Core concepts (Projects, Schema/Content Models, Entries, Assets, Environments, Locales, Stages) map cleanly to standard web-development mental models, and the GraphQL API is auto-generated from the schema so the surface is self-documenting via introspection. Not higher because the 2025–2026 additions (Content Federation Remote Sources, Variants, Taxonomies, Content Workflows, UI Extensions) add real optional conceptual load once teams move beyond a single-project marketing build.

Framework-specific quickstarts (Next.js incl. App Router, Nuxt 3, Astro, SvelteKit, Remix, Angular), in-console GraphQL Playground, sample/starter projects on creation, Hygraph Academy, and an active YouTube channel provide a layered path, reinforced by G2's #1 Implementation ranking in Headless CMS across five consecutive reports. Not higher because Academy is lighter than Contentful Learning Services and there is still no in-product guided tour on par with Storyblok or Sanity.

Consumption is plain GraphQL via standard clients (graphql-request, Apollo Client, urql) with first-class Next.js treatment (App Router / React Server Components), plus documented Nuxt 3, Astro islands, SvelteKit, Remix, and Angular guides — no proprietary SDK or query language to learn. Not higher because GraphQL-only delivery still imposes a small barrier for REST-reflexive teams compared to peers offering both transports.

The vendor's starter set is aging: the canonical hygraph-examples repo was last pushed in July 2024, the Next.js SaaS reference-marketing-website in January 2025, and the Next.js blog starter in August 2024, while the newest Next.js reference (hygraph-showcase, April 2026, TypeScript + codegen + live preview + federation) assumes access to Hygraph's own Showcase project rather than shipping a cloneable schema. Schema Import (August 2026) lets teams copy models, components and remote sources from a template project, which partly offsets this. Not lower because the starters still bundle schema, seed content and typed GraphQL; not higher because breadth and freshness trail Storyblok, Sanity and Contentful.

Baseline integration needs only the Content API endpoint plus a Public Auth Token for delivery (optional Management/Personal Access Token for writes), with straightforward Master vs custom environment selection — meaningfully simpler than Contentful's space/environment/key matrix. Not higher because Content Federation adds non-trivial setup (Remote Source registration, field bindings, schema extensions) and the App Framework for UI extensions carries its own configuration surface.

The first-party Management SDK now provides migration-as-code — typed batch migrations applied in a single transaction, getEnvironmentDiff, and version-controlled migration statements including field renames — a genuine equivalent to contentful-migration that the prior scoring understated, and there is no 50-field-per-type cap. Not higher (still below 60) because renaming fields breaks consuming GraphQL queries downstream and field-type changes on populated models still carry real risk on live content even with tooling.

Live Preview now renders the frontend in a side-by-side iframe inside Hygraph Studio — editors open an entry, click Open live preview, and see unpublished DRAFT-stage changes on the real site without leaving the form — and Click to Edit (via the Hygraph Preview SDK) links preview elements back to editor fields. Not higher because the frontend must still implement preview mode wired to the DRAFT stage and install the Preview SDK, so it is not the fully plug-and-play visual editor Storyblok offers.

GraphQL querying and schema design are the only specializations beyond general web development, and a React/TypeScript developer with basic GraphQL experience can be productive in days with no certification requirement. Not higher because GraphQL remains a marginal skill barrier versus REST-only peers, and Content Federation plus App Framework UI Extensions add depth for advanced builds.

SaaS hosting removes all CMS infrastructure/ops staffing, so a 2–3 person team (frontend developer + content strategist, optional shared technical oversight) can ship a production marketing site, and solo developers handle smaller projects. Not higher because larger projects with Content Federation, multiple environments, multi-locale, Variants, and custom UI Extensions justify 3–5 people — consistent with the headless CMS peer band.

Editors can create variants without schema changes, use AI Assist and AI Agents, jump from the live site into fields with Click to Edit, copy entries across projects (September 2026), and run Content Freeze windows themselves. Not higher because there is still no first-party marketer-facing page builder, so new content types or page patterns remain developer/admin work, and approval workflows and scheduled publishing are Enterprise-only per the pricing page, so Growth-plan teams lack them.

Pure SaaS updated continuously by Hygraph with no customer-run upgrades; the 2025 systems-API redesign was migrated by the vendor, and the monthly changelog (Apr–Sep 2026) is additive features, improvements and bug fixes with no breaking-change entries. Regional cluster maintenance is not invisible, though: 2-hour scheduled windows hit shared-us-west-2 (Oregon) on 2026-04-30 and ap-south-1 (Mumbai) on 2026-04-23. Upgrade work otherwise means opting into new capabilities (AI Agents, Click to Edit, Components) and routine semver SDK bumps.

SaaS on AWS: infrastructure, OS, runtime and application patching are vendor-managed and invisible to customers. SOC 2 Type II and ISO 27001 constrain patch practice, status.hygraph.com discloses incidents, and no Hygraph security advisories or CVEs surfaced for 2025–2026. The customer's only job is keeping its own GraphQL client and renderer packages current through normal npm updates.

Hygraph has forced two customer-facing changes in four years: the 2022 GraphCMS-to-Hygraph rebrand (endpoint and branding updates) and retirement of the legacy content endpoint for existing projects by end of Q1 2025, which required switching to the high-performance endpoint. Both were URL-level switches rather than query rewrites, and the 2025 systems-API redesign was migrated by the vendor. Recent 2026 releases are additive, but the endpoint retirement and short platform history keep this mid-range rather than 70+.

SaaS removes all server-side dependency management. The client-side surface is small: any standard GraphQL client (graphql-request, Apollo, urql) plus an optional rich-text renderer. There are no mandatory Hygraph-specific runtime dependencies, so supply-chain exposure and upkeep stay low.

The project dashboard tracks API operations, asset traffic and storage against plan limits, the webhook Logs view shows delivery status for the last 7 days, and status.hygraph.com covers incidents. The pricing matrix lists Performance Monitoring and Reporting as Enterprise-only, so Hobby and Growth teams need their own APM and alerting for latency and limit proximity; Growth overages bill automatically at $0.20 per 10K operations, which makes that alerting matter.

Required fields and reference validation stop some hygiene problems at write time, Taxonomies help with organization, and AI Agents (Translation, SEO, Summarizer; updated May 2026) cut manual editorial effort on some tasks. There is still no automated orphan-asset detection, broken-reference alerting, content-expiry workflow or content health dashboard, so hygiene depends on editorial discipline. Scheduled publishing and workflows are Enterprise-only, which adds manual work on lower plans.

A CDN-backed content API cache and asset delivery handle read performance globally without customer tuning. GraphQL query design is the main lever, and there is no built-in query complexity scoring or production warnings. Read replicas and custom database indexes for complex queries exist only as Enterprise add-ons, so heavy query workloads on lower plans have no tuning knobs beyond reshaping queries.

The pricing matrix is explicit: Hobby gets community support, Growth (from $199/month) gets community plus an email support desk, and only Enterprise gets priority support, a CSM, onboarding, training and custom support SLAs. Review sites have long praised Hygraph's support responsiveness, and Intercom in-app chat is offered to all, but there is no published response SLA below Enterprise. That puts it in the 'good support requires Enterprise' band.

The community channel is Slack (slack.hygraph.com), which the pricing FAQ names as the main free support route, and Hygraph staff take part there. Official SDK repos under github.com/hygraph get maintainer responses. Slack content isn't publicly indexed and Stack Overflow coverage is thin next to Contentful or Sanity, so answers to niche questions are hard to find without asking, and there is no public roadmap or ideas board.

SaaS delivery puts fixes in front of every customer on deploy, and the changelog carries a dedicated Bug fixes section in each monthly entry (e.g., batch migrations not applying renderer changes, enum app-renderer saves). 2026 incidents were short and resolved quickly: login failures of 54 and 10 minutes on 2026-06-26 and a 35-minute sign-up outage on 2026-08-07. Non-critical requests still move at a mid-size vendor's pace, and with no public roadmap or issue tracker there is little visibility into when they land.

Landing page creation still requires developer-built page templates that content editors populate with structured content. Click to Edit (Visual Editor, public beta as of May 2026) lets editors hover over rendered content, click 'Edit in Hygraph', and jump directly to the corresponding field in Studio, with tight two-way sync and Live Preview support for Next.js/Nuxt. Hygraph also positions a lead-generation/campaign-pages use case where modular blocks let marketers iterate campaign pages without per-iteration developer work. But there is still no drag-and-drop page builder: marketers cannot create new page layouts or new component types without developer involvement. Compared to Storyblok (true visual page building) or Contentful Studio, Hygraph remains in the 'edit existing layouts, not create new ones' band.

Hygraph has no native campaign management: no content calendar, no multi-channel campaign coordination and no campaign analytics. What it does have is launch tooling. Releases group entries for coordinated publishing, scheduled publishing handles timed launches, and Content Freeze (June 2026) lets admins block content changes during critical windows such as a major launch, with a permission-controlled setting and a banner in Studio. Scheduled publishing, Custom Workflows and Variants are all Enterprise-only on the current pricing page, so Growth teams get none of this. Marketing teams still need a separate martech stack to plan and measure campaigns.

SEO metadata is still modeled as ordinary content fields, with no built-in SEO field type or validation, and sitemaps, redirects and robots.txt remain frontend responsibilities. The SEO/GEO Expert Agent runs inside workflow steps and posts an SEO report on the entry before publish; since July 2026 it can also produce Generative Engine Optimization (GEO) review reports. Slug templates can now pull a field from a related model. Agent-driven review inside the editorial workflow lifts Hygraph above plain manual SEO fields, but structural SEO infrastructure (sitemaps, redirects, schema.org) is still code-side, and AI Agents are priced as an add-on (about $10K a year per the pricing FAQ).

No built-in form handling, CTA management, lead capture, or conversion tracking. All performance marketing infrastructure requires frontend integration with external tools (HubSpot Forms, Typeform, conversion tracking pixels). Performance marketing teams will always need a parallel toolset alongside Hygraph. Consistent with the headless CMS architecture philosophy but a significant gap for marketing-led evaluations.

Hygraph's Personalize framework covers what (Variants, multiple versions inside one entry), who (Segments) and how (rules surfaced through the API). Per the current pricing table Variants (up to 30 per entry) and Segments (up to 500) are Enterprise-only; Hobby and Growth get neither. Segments rely on an external CDP for user profiles and the real-time decision runs in the frontend or a personalization engine. The marketplace lists Dynamic Yield, Segment and Optimizely, but those pages link to the vendors' own generic setup docs rather than a Hygraph app. This stays in the 'rule-based targeting or personalization via integration' band.

There is no native A/B testing engine. Variants and Segments can back content experiments, but the experiment runtime, statistical reporting and winner selection have to live in an external personalization engine or CDP, which is how Hygraph's own docs describe A/B testing. The marketplace has an Optimizely listing, but its installation link points to Optimizely's .NET CMS install guide, so it is a partner listing rather than a working Hygraph connector. There is no experiment dashboard or auto-winner inside Hygraph.

AI Agents (Translation, Summarization, SEO/GEO) run autonomously inside Content Workflow steps — meaningfully reducing manual work between draft and publish — and AI Assist provides in-editor generation, translation, and inline refinement. Click to Edit Visual Editor with Live Preview and Next.js/Nuxt support, modular content blocks, and approval workflows on Enterprise tiers all compound editor velocity. Marketing teams can assemble, iterate, and launch campaign pages from existing templates without per-iteration developer involvement. However, new page templates and layouts still require developer build time, keeping Hygraph in the 'adequate speed with some manual steps' band rather than sub-hour brief-to-publish for greenfield pages.

The headless GraphQL architecture natively supports structured content delivery to any channel — web, mobile apps, kiosks, digital signage, and IoT surfaces. Content Federation allows a single Hygraph API endpoint to aggregate content from multiple remote sources for omnichannel delivery. This is a core architectural strength of the platform and aligns with its positioning as content infrastructure. The score is capped at 55 because channel-specific renditions (email rendering, push notification formatting, SMS truncation) require frontend/middleware implementation rather than native channel management.

There are no analytics dashboards or content performance metrics inside the CMS. The marketplace has a Google Analytics listing, but its installation link goes to Google's general developer docs, not a Hygraph app, so analytics are still implemented as frontend tags. The April 2026 Usage dashboard reports API calls, bandwidth and request activity, which is billing and operations data rather than content performance. Marketers cannot see engagement or content decay inside Hygraph.

No brand guardrails at the CMS layer — theming, design tokens, and visual consistency are entirely frontend responsibilities. Content modeling with required fields, validation rules, and component-level structure enforces content quality and structural consistency, but marketers can still deviate from brand voice or add off-brand content without CMS-level enforcement. For structured content accuracy, the platform is good; for brand presentation enforcement, it is entirely frontend-dependent.

No native social publishing workflows, push-to-social automation, or social scheduling. OG/Twitter card meta fields must be modeled as standard content fields with no built-in OG field type or validation. UGC embed support requires custom frontend implementation. Hygraph can enable good social metadata via structured content modeling but requires developers to set it up and provides no out-of-the-box social integration.

Hygraph does have its own asset layer: files are uploaded to a built-in Digital Asset Manager with unlimited storage on every plan, served over a global CDN, with on-the-fly image transformations (the 2026 changelog added safeguards to that transformation pipeline). Installable marketplace apps extend this to Bynder, Aprimo, Cloudinary, imgix and Scaleflex Filerobot for enterprise DAM, and Mux for video, while Remote Fields can surface external DAM assets through the API. What it lacks natively is rights management, usage tracking and rich asset tagging, which still depend on the external DAM. That puts it in the 'media library with transforms' band. The previous assessment wrongly said Hygraph stores only references to external assets.

Variants plus Locales handle locale-specific content without duplicating entries, and editors can now copy a field value from the default locale to every other locale in one action for fields that do not need translation. The Translation Agent localizes content inside workflow steps, alongside installable Crowdin, Smartling, Lokalise, DeepL and EasyTranslate apps for professional translation workflows. Growth includes 3 locales and Enterprise up to 80, with extra locales at about $150 a month each. Market-level scheduling is limited to per-entry scheduled publishing, which is Enterprise-only, and regional compliance content is frontend-side.

Hygraph's marketplace now lists Salesforce Marketing Cloud, Klaviyo, Segment, Dynamic Yield, Zapier and n8n under Marketing, but every one of those pages links its installation guide to the vendor's own generic setup docs, so they are partner listings rather than pre-built Hygraph connectors. There is still no HubSpot or Marketo connector. Real MarTech integration runs through webhooks (10 on Growth, up to 100 on Enterprise), the APIs and the MCP server, which since August 2026 is available as a connector in Claude. That is some integration surface plus generic webhooks, at the low end of the band.

Product content is a legitimate Hygraph use case. Teams model descriptions, variant copy and rich media as content types and federate live product data from commerce platforms into the same GraphQL endpoint, and editors get product pickers from the Shopify, BigCommerce and commercetools apps. A documented Akeneo PIM integration lets product information flow in from a PIM. There is still no purpose-built PIM, SKU management or pricing engine in Hygraph itself, and Remote Sources are limited to 1 on Growth (up to 10 on Enterprise), so this remains generic content types adapted well for product content.

No native merchandising features. Category management, promotional content scheduling, cross-sell/upsell content management, search merchandising rules, and visual merchandising tools are all absent. Hygraph is not positioned for merchandising and provides no tooling for it. Commerce use cases require Hygraph to serve as content layer only, with all merchandising logic in a dedicated commerce platform.

Installable commerce apps cover Shopify (product picker, Hydrogen starter), BigCommerce, commercetools (the deepest, with product, variant and category picking in Studio) and Alokai, and Content Federation can federate any commerce GraphQL or REST API into the Hygraph schema. That is API-level federation plus UI pickers across several platforms. The marketplace's Salesforce Commerce Cloud, SAP Commerce Cloud and Adobe Commerce pages, however, link to the vendors' own generic docs rather than a Hygraph app, so those three are federation-by-hand rather than packaged integrations. Federation is gated by plan (1 remote source on Growth).

Content Federation enables editorial pages that blend rich content with real-time product data (prices, availability from Shopify/commercetools) at query time — supporting buying guides, lookbooks, and shop-the-look patterns via API composition. The Shopify Product Picker allows editors to reference products inline in content entries. However, there is no native 'shoppable content' widget, no inline cart/purchase CTA managed at the CMS layer, and the editorial-commerce blending requires significant frontend implementation. This is a plausible but developer-heavy use case.

No direct checkout or cart CMS capability. Hygraph serves content to the frontend; checkout and cart logic live in the commerce platform. Promotional banners, trust badges, and upsell messaging can be managed in Hygraph as content types and surfaced in checkout pages via the delivery API, but this requires custom frontend integration for every touchpoint. There is no pre-built CMS-managed cart content injection capability.

No native post-purchase content management tied to order events. Order confirmation content, delivery tracking pages, and loyalty program content live in the commerce platform's transactional templates. Hygraph could theoretically serve as a content backend for post-purchase email templates or onboarding sequences via API, but there is no order event integration, no CMS-managed post-purchase sequence manager, and no documented post-purchase use case.

Hygraph participated in B2B Atlanta 2025, signaling a B2B eCommerce content positioning. Multi-project configurations, granular permissions, multi-locale support, and Environments support B2B scenarios with multiple buyer portals or regional storefronts. However, there are no native B2B-specific content features: no gated catalog management, no quote-request flow tooling, no account-based content segmentation, and no spec sheet/documentation management purpose-built for B2B. Access control allows restricting content creation but not buyer-facing catalog gating.

No native commerce search engine or search landing page management. Content is queryable via the GraphQL API with filtering, ordering, and pagination, but faceted search enrichment, synonym management, and blended content-product search results require a dedicated search layer (Algolia, Elasticsearch, Coveo). Content Federation can proxy external search service results through the Hygraph endpoint, but this is a developer architecture pattern, not a commerce search solution.

Promotional content (sale banners, countdown copy, promo messaging) can be modeled as content types, and scheduled publishing plus Releases handle time-based activation. Scheduled publishing is Enterprise-only on the current pricing page, so Growth teams must publish promotions by hand. There are no countdown widgets, channel-specific targeting controls or promo code tooling at the CMS layer.

Multiple storefronts are architecturally supported via separate Environments within a project or separate projects per storefront, with content models shared or independently configured. The dedicated Multi-Brand CMS use case page positions Hygraph for managing large-scale commerce operations with products across markets via flexible content modeling, granular permissions, and internationalization. Content Federation supports hub-and-spoke architectures where shared product content is centralized and storefront-specific editorial is managed per project. However, multi-storefront setup still requires intentional schema design and considerable architectural work; there is no turn-key multi-storefront management UI.

No native 360-degree views, AR/3D model references, or image hotspot management in Hygraph. The Bynder DAM integration and Remote Fields can surface rich media assets managed in Bynder, including video and high-resolution images. Basic image upload and video embed via standard media fields cover standard product imagery needs. Advanced visual commerce experiences (AR try-on, 360° viewers, interactive hotspots) require frontend implementation with external tools.

No marketplace content management features. Multi-author content is technically possible within Hygraph's permission model — different sellers could have restricted roles that only allow them to create/edit their own content types — but this requires custom schema design and RBAC configuration rather than native marketplace tooling. Seller profile management, content quality moderation at scale, review aggregation, and marketplace-specific workflows are all absent.

Variants + Locales handle locale-specific product descriptions and regional content variants without entry duplication. The Translation Agent autonomously translates product content within Content Workflow steps, alongside existing Crowdin, Smartling, and EasyTranslate apps for human-in-the-loop translation workflows. Locale-specific content is surfaced via the GraphQL API with locale filtering. Currency-aware content blocks and EU regulatory labels (Prop 65, ingredient lists) require frontend implementation or custom content type modeling — not natively enforced.

No connection between content engagement and commerce outcomes within Hygraph. Revenue attribution to content pages, content-assisted conversion tracking, and product content performance metrics are entirely absent. Analytics depend on frontend tracking implementations (GA4, Segment, or commerce platform analytics). Hygraph has no awareness of purchase events or downstream conversion data from the CMS layer.

Custom roles with model-level permissions provide granular access control: separate permissions for create, read, update, delete, and publish operations per content type, content stage, locale, and environment. Granular Permissions extend conditional access based on field values in some configurations. Public API Endpoints can be scoped to specific content models and stages. SSO on Enterprise plans. However, audience-based content visibility (show entries to specific departments) is not natively supported, and field-level row-security for sensitive data remains a gap.

Content types can model knowledge base articles, FAQs, and documentation effectively, and Hygraph now maintains dedicated Knowledge Management and Portals use case pages positioning it for centralized, structured, reusable knowledge disseminated to any destination. Taxonomies enable structured classification for content discovery and knowledge navigation. Content Workflows provide approval-based publication governance. Reference case: 2U (edtech) replaced its internal CMS with Hygraph to serve 300,000+ students across LMS, websites, and portals. However, there is no purpose-built knowledge lifecycle (review dates, expiry, archival automation), no stale content flagging, and search quality requires an external search service at scale.

Hygraph markets a Portals use case (including enterprise intranets) but is not designed for employee-facing portal experiences out of the box. No notification system for content consumers, no social features, no employee directory integration, no personalized content dashboards for end-users, and no portal UX components. Building an intranet on Hygraph is technically feasible via the API (as the 2U case study demonstrates for student-facing experiences) but requires building every portal feature from scratch in the consuming application. Teams targeting intranet use cases should evaluate platforms purpose-built for workplace experience.

No native internal communications tooling. There are no targeted announcement features, no broadcast channels, no read receipts, and no acknowledgment tracking. Company news or department announcements can be modeled as content types and published via workflows, but this is a basic publishing pattern with no audience targeting, no mandatory-read enforcement, and no delivery channel integration. Purely a content storage-and-delivery capability, not an internal comms platform.

No native people directory, org chart, or HR system integration. Employee profiles could theoretically be modeled as content types with relationship fields for manager hierarchies, but this is a custom build entirely. No Workday, BambooHR, or HRIS connector is documented. No org chart visualization component exists at the CMS layer. Hygraph provides zero out-of-the-box directory functionality.

Policies and SOPs can be modeled as structured content with version history: 14 days of retention on Growth and up to 365 days on Enterprise. Audit logs (Enterprise, 90-day retention) give change tracking, and Custom Workflows (Enterprise) route policy approvals. There is no acknowledgment tracking, no review or expiry reminders and no policy templates. Adequate for small-scale policy publishing, not purpose-built for policy management.

Onboarding content paths can be structurally modeled as content types with role references, but Hygraph provides no native progressive disclosure engine, no task checklist management, and no HR-triggered new-hire journey automation. The 2U case study demonstrates Hygraph being used for educational content sequencing, which is architecturally analogous to employee onboarding, but in both cases the sequencing and progression logic must be built in the frontend application. Basic onboarding pages are achievable but the journey management layer is entirely custom.

No native enterprise search engine. Internal content is queryable via the GraphQL API with filtering and ordering, but faceted filtering, AI-powered relevance ranking, and federated cross-system search require a separate search layer (Algolia, Elasticsearch). The MCP Server enables LLM-driven content discovery and Q&A patterns over the Hygraph content graph — a modern alternative to traditional search for AI-native intranets — but does not constitute a packaged enterprise search experience. Content Federation can proxy external search results through the Hygraph endpoint. Search quality for large intranet content volumes without an external search service remains poor.

The Hygraph Studio CMS interface is web-based and responsive but there is no dedicated mobile editor app. Frontline content delivery is via the headless API — mobile consumer experiences depend entirely on the implementing team's frontend app. No native offline support, push notifications for content consumers, or kiosk/shared-device modes exist at the CMS layer. Responsive web access for editors is the extent of mobile capability.

Hygraph can serve as a content backend to LMS platforms via its GraphQL delivery API, as demonstrated by the 2U case study where Hygraph powers content delivery to 300,000+ students across LMS portals and websites. However, there is no native LMS connector, no SCORM/xAPI support, no course assignment management, and no completion tracking. Integration with LMS systems (Cornerstone, Workday Learning) requires custom API development. Content hosting is capable; learning infrastructure is absent.

No social or collaboration features for content consumers. No comments, reactions, discussion forums, peer recognition, polls, idea submission, or community spaces. Real-time collaboration in the Hygraph Studio editor is limited compared to platforms like Sanity. Hygraph is purpose-built as content infrastructure, not a social layer. All employee engagement features would require a separate platform or entirely custom frontend development.

No documented native integrations with Microsoft Teams, SharePoint, Slack, or Google Workspace. This is a significant gap for intranet scenarios where content surfacing via Teams cards, Slack notifications for new content, or embedded Google Drive documents is expected. Hygraph's webhook infrastructure could theoretically trigger Slack or Teams notifications when content is published, but this is developer-configured custom integration, not a pre-built workplace tool connector.

No automated review dates, stale content flagging, or archival workflow automation. Content versioning history (30-day minimum) allows rollback but not proactive lifecycle management. Audit logs on Enterprise track who changed what but do not flag aging content for review. Content Workflows can enforce approval for publication but do not have time-based expiry triggers. For intranet content governance (policy expiry, outdated documentation), manual processes are required.

No internal content analytics, department-level views, failed search term reporting, or adoption dashboards within the CMS. Content delivery metrics, if tracked at all, live in external analytics tools (GA4, Segment) configured in the consuming frontend application. Hygraph provides no visibility into which employees are reading which content, what intranet content is underperforming, or how content adoption is trending across departments.

Hygraph is logically multi-tenant with physically shared infrastructure for standard plans. The Enterprise plan explicitly includes multi-tenancy support that maps to organizational structures 1:1, with dedicated infrastructure available as an Enterprise option. Each project has independent content models, environments, API tokens, and roles — providing genuine content and schema isolation. Organization-level management allows centralized billing and user administration across projects. Content Federation enables selective data sharing between projects without breaking isolation.

Hygraph's Multi-Brand CMS positioning explicitly references shared content models that every brand inherits — central teams define the structure and governance once and replicate it across brands, markets, and channels. In practice, this is implemented via Content Federation (hub-and-spoke with Top-Level Remote Fields) plus schema templates rather than true cross-project schema inheritance, so it is still federation-based sharing rather than native shared component libraries at the schema layer. The marketing message is stronger than the underlying mechanism, but the federation patterns themselves have matured.

Organization-level administration provides centralized user management, SSO, and billing oversight. Project-level roles allow brand-specific permission management. Content Workflows enable approval-based publication governance within each project. AI Agents and the MCP Server are governance-aware and operate strictly within existing roles and permissions structures. However, cross-project governance — enforcing brand standards across projects, cross-brand approval workflows, centralized content policy enforcement — is not natively supported and requires process-based governance rather than platform enforcement.

The self-serve tiers are now just Hobby (free) and Growth ($199/month per project); the Professional and Scale tiers are gone. Multitenancy, SSO and custom roles are Enterprise-only, so a real multi-brand deployment means Enterprise, which the pricing FAQ puts at about $32,000 to $300,000 a year, built as a platform fee plus one or more projects per use case. The shared platform fee gives some economies over buying brands separately, but each added brand project still adds cost, so scaling is roughly linear.

No brand theming at the CMS layer. Per-brand visual identity — design tokens, typography, color palettes, logo treatment — is entirely the responsibility of the frontend application consuming Hygraph content. Content models provide structural consistency for content types per project, but brand presentation is decoupled from the CMS by design. Multi-brand deployments must implement brand theming entirely in their frontend code.

Each project can have independent locale configurations and translation workflows, allowing per-brand localization management. The autonomous Translation Agent and the Crowdin/Smartling/EasyTranslate apps are configurable per project, enabling brand-specific translation approval flows. However, shared vs. isolated translation workflow governance across brands — e.g., a global legal disclaimer translated once and pushed to all brand locales — is not natively managed and requires Content Federation patterns plus custom orchestration. Regional compliance content (cookie consent, legal disclaimers) is frontend-side.

No cross-brand analytics dashboard within Hygraph. Each project is a discrete entity with no aggregated reporting view across the brand portfolio. Publishing cadence, content freshness, and content velocity metrics would require manual data collection from each project or custom API-based aggregation. Portfolio-level content performance measurement is entirely absent from the platform.

Content Workflows are configurable independently per project, meaning each brand can have distinct approval chains, review stages, and publishing rules. Enterprise organizations can maintain centralized audit visibility across projects via Organization-level audit logs. AI Agents operate within each project's governance rules. However, there is no central workflow template management (changes to a shared workflow template don't propagate to brand projects automatically) and cross-brand workflow comparison is manual.

A global content project can feed brand projects through Content Federation and Top-Level Remote Fields, so corporate content appears as a field in each brand's schema. Since September 2026 editors can also copy a saved entry straight into another project and environment as a draft, with field-matching rules that copy matching fields and locales and skip references. That adds a practical copy-then-adapt path for press releases or disclaimers. Copies do not stay linked, though: there are no controlled override points and no push of later corporate updates to brand copies.

No per-brand compliance rules with publishing guardrails within the CMS. Basic compliance settings — GDPR data residency options on Enterprise, audit logs for regulatory evidence — are available at the organization level. Accessibility standards enforcement, regional cookie policies, and data residency per brand project must be managed through process governance rather than platform guardrails. Publishing is not blocked for non-compliant content by default.

No centrally maintained design system at the CMS layer. Schema templates in the marketplace enable some cross-project model replication but are point-in-time copies rather than live inheritance. Updating a 'base' component model does not propagate to brand projects. Brand-level extensions of shared base models are not natively supported — each brand project maintains its own independent schema. Frontend design systems (Storybook, Tokens Studio) are entirely outside Hygraph's scope.

A user can have different roles across multiple brand projects — central admins can be granted full access across all projects while brand teams manage their own users independently. Organization-level management provides a single view of all members and their project assignments. SSO on Enterprise enables unified authentication across all brand tenants. However, cross-brand contributor roles (a user with the same editorial role across all brand projects) require manual per-project assignment rather than a single cross-brand role definition.

Each brand project has its own content model. Schema Import (August 2026) lets a team copy models, components, remote sources, enumerations and taxonomies from another project or template into the current one without rebuilding them, renaming incoming elements on conflict rather than touching existing schema. That makes rolling a base model out to new brands much faster. It is still a one-time copy: later changes to the base do not propagate, and a brand cannot extend a shared base model without forking it. Multi-brand modeling remains copy-and-diverge.

No portfolio-level reporting within Hygraph. There are no executive dashboards showing content freshness by brand, publishing SLA adherence across the portfolio, cost allocation per project/tenant, or capacity planning metrics. Organization-level audit logs provide a governance trail but not portfolio performance reporting. Multi-brand operators must aggregate metrics manually or build custom reporting tooling via the Hygraph API.

Hygraph GmbH is Berlin-headquartered, states GDPR compliance, offers Frankfurt hosting on every plan (EU is one of the 3 shared regions even on Hobby/Growth), and its own FAQ says sub-processors are documented in its Data Processing Agreement and that services outside the EU can be restricted. However, the DPA is not published (hygraph.com/legal and /legal/dpa return 404), there is no public sub-processor page, and data-subject rights are handled by emailing support. Not higher because the DPA and sub-processor list are not publicly available for review and erasure tooling is API/support-driven; not lower because a DPA exists, EU residency is available on all tiers, and the controller is an EU entity.

Hygraph's own published FAQ states it 'does not handle HIPAA-type information' and offers to share its SOC 2 report instead; no BAA is offered and there is no healthcare deployment guidance anywhere on the site. AWS hosting does not make the platform HIPAA-covered. Not higher because the vendor explicitly positions itself outside HIPAA scope; not lower because non-PHI healthcare marketing content remains a supported use.

Coverage rests on GDPR plus regional hosting in the UK, Canada, Brazil, India, Japan and Australia, which helps customers meet local data-handling rules, and Hygraph asserts its annual SOC 2 audit supports DORA requirements. There is no FedRAMP, IRAP, C5, PCI-DSS or HITRUST, and because the DPA is not public its CCPA, UK IDTA and LGPD coverage cannot be verified. Not higher because no government or industry framework is held and regional-law coverage is unverifiable; not lower because broad in-country hosting and a GDPR/DORA posture exceed a GDPR-only baseline.

Hygraph has held SOC 2 Type 2 since 2023, audited by Barr Advisory, and the announcement (updated September 2026) confirms the audit was completed 'for another year' covering security, availability and confidentiality, with a 'Request Report' path for customers. Not higher because the auditor statement in the same post references only the Security criterion and the report is not self-serve through a trust portal; not lower because an annually renewed Type 2 with three TSCs is confirmed.

The security page states Hygraph 'offers hosting on ISO 27001 certified infrastructure', and the content-governance page lists ISO 27001 alongside SOC 2, but no certificate, registrar or scope statement shows a Hygraph-held ISMS certification, and ISO 27018 is not claimed. Not higher because the only precise wording scopes ISO 27001 to the hosting infrastructure and there is no ISO 27018; not lower because ISO 27001 coverage is consistently stated.

Beyond SOC 2 and ISO 27001, Hygraph holds no further certifications: no CSA STAR, PCI-DSS, C5 (notable for a German company), Cyber Essentials Plus, FedRAMP or IRAP. It does run an annual external penetration test and shares the report with customers, which supports procurement but is not a certification. Not higher because there are no additional certifications to add on top of the base; not lower because the annual pen-test report is a real procurement artifact.

Enterprise projects can choose from 9 shared-cluster regions across North America, Europe, LATAM and APAC, or 34 regions on a dedicated cluster, and the vendor can restrict services outside the EU; Hobby and Growth get 3 shared regions (EU, US, APAC). Not higher because the global CDN caches delivered content outside the chosen region, the broad choice is Enterprise-gated, and moving regions requires cloning the project; not lower because the region range far exceeds an EU/US binary.

Content, assets and schema can be extracted and deleted through the GraphQL Content and Management APIs, and Enterprise can add 7-day automatic backups. However, the public Terms make the customer 'solely responsible for exporting' content before termination and promise only a 'reasonable opportunity' to retrieve it, with no published post-termination retention or deletion period, and there is no self-service DSR tooling. Not higher because retention after termination is not publicly documented; not lower because API export and deletion are complete.

Enterprise audit logs capture schema and content changes and admin actions, attribute them to UI users, PATs, app tokens or public API calls, and are filterable with a payload detail view; retention is 90 days (30 on some SLA tiers). The documentation describes only the in-app Audit Logs Explorer, with no export, API endpoint or SIEM integration. Not higher because there is no documented export or SIEM path and retention is short and not customer-configurable; not lower because coverage and actor attribution are comprehensive and retention is published.

Hygraph publishes no accessibility statement, WCAG target or ATAG commitment for its authoring app: there is no accessibility page (hygraph.com/accessibility is 404) and none of the site's 1,434 sitemap URLs or the docs llms.txt address editor accessibility. The earlier assumed WCAG 2.1 AA target could not be found in any vendor source. Not higher because there is no documented commitment for the editor; not lower because the app is a modern React UI with standard keyboard operation for basic field editing.

There is no VPAT, ACR, Section 508 statement, ATAG assessment or accessibility page of any kind, so buyers with procurement accessibility requirements have nothing to review. Not higher because no accessibility documentation exists; not lower because a VPAT could plausibly be obtained via the Enterprise IT security review process, though none is advertised.

AI Assist (GA, all plans with per-plan token allowances) is schema-aware and embedded in the content editor, offering Create, Improve, and Translate tasks across text, Markdown, and rich text fields with a review panel before anything is applied. Project-level AI Guidelines (Brand Voice & Tone, Audience & Context, Legal & Compliance, Glossary) can now be assigned to AI Assist tasks as well as agents, and the Content Summarizer agent adds bulk generation across up to 50 entries with field-by-field review. Not higher because there are no reusable custom prompt templates and generation stays limited to text fields.

Hygraph has no native AI image generation or built-in alt-text automation. Alt-text generation is available exclusively via AltText.ai, a third-party marketplace Technology Partner app that adds an 'Alt Text' sidebar button to the Asset editor and calls the AltText.ai external API (requires a separate AltText.ai account and subscription). No AI video processing, smart crop, or native media AI within the DAM was found.

AI Assist (GA, all plans) translates selected fields into one or more configured locales with optional regional style prompts, and the Translator agent (Enterprise) translates in bulk across up to 50 entries or automatically at a workflow step, with side-by-side field-level review. Glossary guidelines hold per-locale term groups that the Translator agent applies directly, alongside custom instructions for formality and untranslated terms; DeepL and Lokalise are available as real marketplace apps. Not higher because there is no per-locale quality scoring and automated bulk translation is Enterprise-only.

The SEO & GEO Expert agent (Enterprise) reviews all entry fields including nested components and posts a structured report on content optimization, heading structure, and metadata gaps; GEO (generative engine optimization) reports were added in July 2026. Metadata can be generated through AI Assist (meta descriptions, alt text strings in text fields) and the Content Summarizer can write excerpts to a target field, but the SEO agent itself only comments rather than writing metadata. Not higher because there is no automatic SEO title/description/tag population, no image-aware alt text, and no on-page scoring dashboard.

The 2026 AI Agents V2 update lets the Translation, SEO Expert, and Content Summarizer agents run in bulk across up to 50 entries directly from the content table, with every change held for human review as a before-and-after comparison and accept/reject/revert available field by field — genuine bulk content operations woven into editorial. Multiple agents can be chained at different workflow lifecycle steps, all logged and RBAC-scoped. Not higher because there is no native auto-tagging, smart scheduling, or duplicate detection outside agent-triggered runs, and the capability is Enterprise-only with just three agent types.

Hygraph ships three named, no-code configurable agents (Translator, SEO & GEO Expert, Content Summarizer) that run manually on up to 50 entries or automatically at workflow steps, with read-only review states, field-level approve/revert, and guardrails preventing deletes or schema changes. Since August 2026 manually triggered agents can also be listed, launched, and monitored from Claude or other assistants through the MCP server, enabling natural-language orchestration with output still landing in draft for review. Not higher because the agent catalogue is limited to three types, agents are an Enterprise add-on (about $10K a year), and there is no custom agent logic or marketplace.

No native content gap analysis, topic clustering, AI-driven performance scoring, content health metrics, stale content detection, or editorial priority recommendations are documented. The Variants and Taxonomies features enable structured personalization segmentation but are content model tools, not AI-driven intelligence. No content intelligence dashboard or trend analysis capability exists in Hygraph's AI stack.

The SEO & GEO Expert agent (Enterprise) acts as a pre-publish audit, posting a structured report covering optimization, heading structure, and metadata gaps, and can run on up to 50 entries at a time or at a workflow step. Legal & Compliance and Brand Voice guidelines shape agent and Assist output, but they steer generation rather than scoring existing content for compliance. Not higher because there is no readability scoring, accessibility auditing, duplicate/thin-content detection, or multi-dimensional quality score at scale.

Hygraph has no native vector search, semantic search, or embedding generation capability. The GraphQL Content API provides structured relational content that can serve as a source for external RAG pipelines (developers fetch content to embed in external vector stores), and the MCP Server enables schema-aware content queries by AI agents. However, neither constitutes built-in semantic search — all AI-native search capabilities require fully custom external integration with third-party vector databases.

Hygraph personalization combines Variants (what to deliver), Segments (who, defined via an external CDP), and Personalization rules (how/when) — a structural schema and rules approach, not a native AI/ML engine. No native model performs variant selection, predictive segment assignment, next-best-content recommendations, or behavioral scoring; any ML 'personalization engine' is an external CDP/engine integrated via API. No native AI personalization layer is present.

Hygraph's official MCP server is now GA (per Hygraph's September 2026 announcement) and was listed in the Claude Directory in August 2026, with a global OAuth endpoint (mcp.hygraph.com/mcp) using the user's own permissions plus project-scoped PAT endpoints. It supports schema inspection and creation, content read/create/update/publish, bulk updates, cross-environment moves, agent runs, and loading AI guidelines, on all plans; a read-only token scope was added in June 2026 and delete/unpublish are blocked by design. Not higher because destructive operations are excluded and workflow-bound agents cannot be triggered via MCP.

Hygraph offers no BYOK or BYOM for its in-platform AI: AI Assist and AI Agents run on Hygraph-managed LLMs with no setting for supplying OpenAI, Gemini, Azure OpenAI, or self-hosted model keys or endpoints. Hygraph's marketing describes the MCP server as 'bring-your-own LLM', but that only means choosing which external assistant operates on content, not powering native features with your own model. Enterprise customers cannot configure a custom model endpoint.

Developer AI tooling centres on the GA MCP server with a published tools reference covering content, schema, guidelines, and agent tools, backed by a strongly typed GraphQL schema that suits RAG ingestion, plus webhooks and the Management SDK for external agent triggers. Docs offer 'View as Markdown' and 'Copy for LLMs' on every page, and there are official guides for Cursor, TypeScript + MCP agents, and Claude-driven migrations and localization. Not higher because there is no dedicated AI SDK, no official LangChain/LlamaIndex/CrewAI connectors, and no native embedding or RAG endpoint.

AI governance is a core design principle: AI Guidelines in four categories (including Legal & Compliance and Glossary) apply consistently to AI Assist, agents, and MCP clients; agent output lands in a read-only review state with side-by-side diffs and field-level approve/revert; agents and MCP cannot delete or unpublish; and RBAC, a read-only MCP token scope, and Enterprise audit logs cover human and agent actions. Not higher because hallucination detection and confidence scoring are absent, there is no documented IP indemnification or AI data-use policy, and audit logs are UI-only with 90-day retention.

Hygraph provides per-project AI token metering on the Billing page (against plan allocations), an AI Tokens Trend chart, a Duration trend view, and a Workflow Runs table listing each agent run's status, result, duration, and start date for investigating failed or unexpected runs. This covers operational cost visibility and run-level monitoring. Not higher because there is no per-user or per-team consumption breakdown, no LLM-level observability (confidence/hallucination flagging, prompt/response review), and no prompt-effectiveness or quality-trend analytics.