Granular Access Control and Role-Based Permissions in KeystoneJS

When building a headless CMS on KeystoneJS, you often need to split access between editors, moderators, and admins, and hide internal notes from regular users. A typical solution is a multi-level access control system: at the operation (CRUD), item, and field level. We have helped more than 30 projects set up such models turnkey—from analysis to deployment, with security and performance guarantees. In one project for a publishing house, we needed 7 roles with different access to 15 lists. The role model via the database allowed changing permissions without restarting the server, cutting change approval time by 40%.

Why KeystoneJS Wins Over Strapi in Access Flexibility

Strapi uses rigid roles with fixed permissions, and Directus offers only three levels (public, readonly, full). KeystoneJS provides four levels—Operation, Filter, Item, Field—plus the ability to store roles in the database. This lets you implement any business rules without hardcoding. For example, we set up a system where an editor can edit only his own drafts, a moderator can publish others', and an admin can delete. The setup took 3 days, whereas on Strapi we would have had to write custom middleware, extending the timeline to 2 weeks—that's 4.5 times longer.

How to Configure Access Control for Multiple Roles in KeystoneJS

Let's break down the setup using a typical project example. First, define the lists and roles. For each role, create a record in the Role list with flag fields as shown below. Then, in the access functions of each list, check these flags. This approach is flexible and allows managing permissions through the admin panel.

KeystoneJS Access Levels: From Operations to Fields

KeystoneJS's access control is built on four levels. Each solves a specific task.

These levels are thoroughly described in the KeystoneJS Access Control Guide.

Here's a combined configuration example for the Post list:

access: {
  operation: {
    query: ({ session }) => !!session,
    create: ({ session }) => session?.data?.role?.canManagePosts,
    update: ({ session }) => ['editor', 'admin'].includes(session?.data?.role),
    delete: ({ session }) => session?.data?.role === 'admin',
  },
  filter: {
    query: ({ session }) => {
      if (session?.data?.role === 'admin') return true;
      return { author: { id: { equals: session?.data?.id } } };
    },
  },
  item: {
    update: async ({ session, item }) => {
      if (session?.data?.role === 'admin') return true;
      return item.status === 'draft' && item.authorId === session?.data?.id;
    },
  },
  fields: {
    internalNotes: text({
      access: {
        read: ({ session }) => session?.data?.role === 'admin',
        create: ({ session }) => session?.data?.role === 'admin',
        update: ({ session }) => session?.data?.role === 'admin',
      },
    }),
    salary: integer({
      access: {
        read: ({ session }) => ['admin', 'hr'].includes(session?.data?.role),
        update: ({ session }) => session?.data?.role === 'admin',
      },
    }),
  },
},

How to Store Roles in the Database

Instead of hardcoding roles in code, store permissions in the database. This allows changing permissions via the admin panel without restarting.

Comparison of approaches:

// lists/Role.ts
export const Role = list({
  access: {
    operation: {
      query: allowAll,
      create: ({ session }) => session?.data?.role === 'admin',
      update: ({ session }) => session?.data?.role === 'admin',
      delete: ({ session }) => session?.data?.role === 'admin',
    },
  },
  fields: {
    name: text({ validation: { isRequired: true }, isIndexed: 'unique' }),
    canManagePosts: checkbox({ defaultValue: false }),
    canManageUsers: checkbox({ defaultValue: false }),
    canManageRoles: checkbox({ defaultValue: false }),
    canPublish: checkbox({ defaultValue: false }),
    users: relationship({ ref: 'User.role', many: true }),
  },
});

// auth.ts
sessionData: 'id name email role { canManagePosts canManageUsers canPublish }',

Usage in the Post list:

access: {
  operation: {
    create: ({ session }) => !!session?.data?.role?.canManagePosts,
    update: ({ session }) => !!session?.data?.role?.canManagePosts,
    delete: ({ session }) => !!session?.data?.role?.canManagePosts,
  },
},

How We Set Up Access: Process and Scope of Work

1. Object and role analysis — define lists, operations, fields. Output: permission matrix.
2. Design role model — create Role list with checkboxes for each permission.
3. Implement access functions — write Operation, Filter, Item, Field for each list.
4. Test scenarios — verify permissions for each role (up to 20 cases).
5. Deploy and monitor — deploy to server and monitor logs.

Common Mistakes in Access Configuration

Beginners often confuse levels—for example, using Filter Access when Item Access is needed. Filter Access works automatically at the query stage and cannot check item fields that aren't loaded yet. Item Access is suitable for checking status or authorship. Another typical mistake is forgetting to include the necessary role fields in sessionData. Without that, access functions won't see the permissions. We test all scenarios to avoid such issues.

Timelines and Cost

A typical role model setup (3–4 roles, 5–10 lists) takes 2–4 days. Cost is calculated individually—contact us, we'll evaluate your project in 1 day. Get an engineer's consultation for your project.

Experienced engineers with 5+ years of work with KeystoneJS will help implement even complex access scenarios. Contact us for a free estimate.

Headless CMS: Strapi, Directus, Sanity, Contentful, Drupal

Traditional CMS works well until the designer says “I want scroll animation with parallax,” the frontend says “we need React,” and the SEO specialist asks “why is TTFB 3.4 seconds?” At that point, monolithic architecture starts to hinder everyone. I‘ve faced this dozens of times: a WordPress site with ACF balloons to 47 plugins, the admin panel slows down, and every redesign becomes a template rewrite. Headless CMS separates content management from presentation. Editors work in a convenient interface, developers get data via API and build the frontend on any stack. Sounds simple. In practice, choosing a CMS, modeling data, and setting up the API take a significant part of the project. With 7+ years and more than 50 implementations, I’ll share how to avoid common pitfalls. Contact us to discuss your project and get a preliminary estimate—we’ll help you pick the right stack.

What are the key benefits of headless CMS over monolithic?

Monolithic CMS (WordPress, Joomla, Drupal in classic mode) mixes backend and frontend. Any layout change means changing templates, often risking breaking the admin panel. Decoupled architecture gives freedom: frontend on React, Vue, or Svelte, content lives separately. Result: improved load speed (LCP often drops from 4–6 s to 1–1.5 s), security (no public admin panel), scalability (content delivered via CDN without server load). Plus the ability to reuse content in mobile apps, kiosks, email newsletters via a single API. A client we recently helped saw LCP improve from 6.2 s to 1.1 s — a 5.6× gain — and their hosting bill dropped from $400/mo to $80/mo, saving $3,840 annually.

How to choose a headless CMS for your project?

No universal tool exists. The choice depends on team, content complexity, and infrastructure. Let’s break down the key options.

Strapi — open-source, self-hosted, Node.js. Suitable for teams needing data control and API customization. Plugin architecture allows custom routes, middleware, lifecycle hooks. REST and GraphQL out of the box. Deploys in about an hour — three times faster than Drupal. Weakness: versions v4 and v5 are incompatible, migration is painful. Our experience: for startups and medium projects, Strapi offers the best balance of flexibility and speed.

Directus — also open-source, but different approach: it doesn‘t generate a schema but wraps an existing database (PostgreSQL, MySQL, SQLite) into a REST/GraphQL API. If you already have a database, Directus connects without migrations. Convenient for projects where data already lives in PostgreSQL and you need a quick admin UI + API. Saves up to 30% integration time.

Sanity — cloud CMS with real-time editor. Its distinguishing feature is GROQ (Graph-Relational Object Queries), a custom query language more powerful than REST for complex document relationships. Portable Text for structured content. Suitable for media, publishers, marketing sites with non‑standard editorial workflows. Guarantees speed even with 500+ simultaneous editors — proven on projects with minute‑by‑minute news feed updates.

Contentful — enterprise cloud CMS. Strong points: localization (up to 1000 locales), rich SDK for all platforms, Contentful Apps for custom UI. Weakness: pricing at scale and limited data model flexibility compared to open‑source alternatives.

Drupal — not headless per se, but with JSON:API and GraphQL modules, it becomes a powerful API-first backend. Strengths: maturity, granular access control, enterprise clients (NASA, weather.com). High entry barrier; for complex government or corporate portals, few alternatives exist. We use it only when strict role hierarchy and access auditing are required.

Consequences of poor content modeling

Content modeling is critical. Mistakes at this stage are costly. A typical problem: a body field of type rich text for everything. Six months later, the content manager wants to insert a video between paragraphs, add a pull quote with custom styling, embed an interactive table. Rich text can‘t handle that. Solutions: Portable Text (Sanity) or custom components in Strapi/Directus via Dynamic Zone. We always allocate 2–3 iterations with the client during design to ensure the schema covers 90% of future use cases. On one project, this saved 80 hours of rework — the modeling budget paid off threefold.

How we build projects on headless CMS

Frontend for headless CMS almost always uses Next.js (App Router) or Nuxt. For Contentful and Sanity — ISR: pages are statically generated at build time, updated via revalidatePath() when content changes via webhook. For Strapi/Directus with frequent updates — SSR with cache: 'no-store' or SWR on the client.

Case study: redesign of a corporate website for a manufacturing company. Previous site: WordPress with ACF, 200+ pages, 4 languages. Problems: TTFB 3.8 s, editors complained about slow admin. Migrated to Strapi (self-hosted, PostgreSQL), Next.js App Router. Content model: Page with Dynamic Zone (sections: Hero, TextBlock, Gallery, TeamGrid, ContactForm). Localization via Strapi i18n plugin + next-intl on frontend. Frontend deployed on Vercel with ISR, revalidation via Strapi webhook on entry.publish. According to the client: TTFB dropped from 3.8 s to 180 ms (static with CDN) — a 21× improvement. Editors got a clean interface without 47 plugins. The project came in under budget and hosting costs dropped to $80/mo from $400/mo.

Implementation process broken into stages:

1. Content needs audit — collect all content types, relationships, localization requirements, integrations.
2. Data schema design — create models, fields, validation, access roles. Document in Swagger/OpenAPI.
3. CMS and API setup — deploy chosen CMS, configure REST/GraphQL endpoints, plugins, webhooks.
4. Frontend development — connect Next.js/Nuxt, configure ISR/SSR, section components, routing.
5. Content migration (if legacy) — automated loading via API or scripts.
6. Testing — check API endpoints, regression, load testing, Core Web Vitals.
7. Deployment — configure CDN, SSL, CI/CD, monitoring.

How long does implementation take?

The standard path includes all stages. Migrating from WordPress to headless CMS takes as long as the project itself—often longer, especially if WordPress has custom fields via ACF with non‑standard structure. Our typical timelines:

Cost is calculated individually after a brief. Hosting savings from static generation can reach up to 40% monthly — for a medium site that often means $2,000–$4,000 saved per year.

Non‑obvious considerations when choosing a headless CMS

• Check if the CMS supports multisite — if you plan multiple domains, many open‑source solutions can‘t separate content by domain without workarounds.
• Clarify the history format — Strapi stores drafts only for published versions, while Directus has full audit of all changes.
• Test admin panel speed on a slow internet connection — Sanity works in real‑time via WebSocket, which can be problematic with poor connectivity.
• Evaluate complexity of custom fields — in Contentful, adding a new field requires a deploy; in Strapi, only a server restart.
• Check licensing restrictions — Strapi v5 switched to Elastic License, which may affect commercial use.

What is included

• Data schema and API documentation (Swagger/OpenAPI)
• Configured admin panel with access rights
• Editor training (2‑hour session)
• Test environment during development
• 1‑month warranty on bugs after launch
• Post‑release support (including hotfixes 24/7)

Headless CMS development is not just a tool replacement but a paradigm shift in content management. We help make this transition without downtime or data loss. Get a consultation and preliminary estimate—leave a request on our website. Order headless CMS implementation with guaranteed results.