# KeystoneJS to Directus migration > How to migrate from KeystoneJS to Directus, with a side-by-side comparison and FAQs. --- ## Source: KeystoneJS - **Type:** headless - **Official site:** https://keystonejs.com - **Last reviewed:** 2026-07-28 ## Key advantages KeystoneJS is one of those tools that really clicks if your team thinks in code. It's a schema-driven, open-source headless CMS built on Node.js, and the developer experience is genuinely good. You define your content models in TypeScript, Keystone generates a GraphQL API and an admin UI for you, and you're off. There's very little magic or abstraction hiding what's happening under the hood, which we appreciate when building complex projects for clients. The Prisma ORM integration is a real highlight. Automatic migration generation, type-safe database access, and support for PostgreSQL, MySQL, and SQLite mean you're not fighting your data layer. If you've ever had to wrangle a CMS into supporting a non-trivial relational content model, you'll understand why this matters. Keystone lets you express those relationships cleanly and query them with a proper GraphQL API. The document field editor is also worth mentioning. It's one of the more thoughtful rich text implementations we've seen in a headless CMS. You can embed custom React components directly into the editor, which means content teams can work with your actual design system components rather than generic blocks. For teams that care about structured content, Keystone gives you real tools to enforce it. Where Keystone really shines is in projects where the development team wants full ownership of the stack. There's no vendor lock-in, no proprietary query language, and no surprise pricing tiers. If you want a CMS that feels like a well-designed library rather than a platform, Keystone delivers on that promise. ### Schema-as-code with full TypeScript support Define your entire content model in TypeScript with strong type inference throughout. The schema drives everything from the database to the admin UI to the GraphQL API. ### Automatic GraphQL API generation Every content type you define automatically gets a full CRUD GraphQL API with filtering, pagination, and relationship resolution. No manual endpoint wiring needed. ### Prisma-powered database layer Built on Prisma ORM with automatic migration generation and type-safe queries. Supports PostgreSQL, MySQL, and SQLite out of the box. ### Flexible document field editor The rich text editor supports custom embedded components that map to your design system, giving content editors structured authoring without sacrificing flexibility. ### Granular access control Fine-grained, field-level access control defined in code. You can write custom logic for create, read, update, and delete operations per field or per list. ### Fully open source with no vendor lock-in MIT licensed with no paid tiers or proprietary features gated behind a subscription. You own the entire stack and can host it wherever you want. ## Key pain points The biggest challenge with KeystoneJS is that it hands you all the responsibility that a managed CMS would normally handle. Deployment is entirely on you, and the documentation around production hosting, Docker configuration, and scaling is thin. We've seen teams struggle to go from a smooth local development experience to a reliable production setup, especially if they don't have dedicated DevOps support. The admin UI Docker image alone can balloon to over a gigabyte, which is a headache for containerised deployments. The community around Keystone is significantly smaller than competitors like Strapi or Payload. That means fewer tutorials, fewer plugins, and slower answers when you hit an edge case. The ecosystem of ready-made integrations is almost non-existent, so you'll be building most things from scratch. For an agency working on client projects with deadlines, that time cost adds up quickly. Content editors also tend to have a harder time with Keystone compared to more polished alternatives. The admin UI is functional but feels utilitarian, and non-technical users often need more onboarding than you'd expect. There's no visual editing, no preview infrastructure, and no real content workflow features like drafts, publishing schedules, or approval chains without building them yourself. If your client's content team needs a CMS they can pick up and run with, Keystone usually isn't the answer. ### Deployment complexity Self-hosting is the only option, and the docs don't hold your hand. Getting Keystone into production requires real infrastructure knowledge, and the large Docker image sizes make it worse. ### Small community and ecosystem Compared to Strapi or even Payload, the community is much smaller. Fewer plugins, fewer tutorials, and slower support when things go wrong. ### No built-in content workflows There are no turnkey drafts, scheduled publishing, or approval chains. Keystone provides field primitives that can be assembled into publishing workflows, but you need to wire them up yourself. ### Admin UI feels dated The admin panel is functional but lacks the polish and UX of modern CMS interfaces. Non-technical editors often find it confusing and need more training. ### No visual editing or live preview There's no way for editors to see content in context before publishing. You'd need to build your own preview infrastructure, which is a significant engineering effort. ### Scaling requires significant effort Running Keystone under high traffic means managing session stores, reverse proxies, and server resources yourself. It doesn't scale as smoothly as cloud-native CMS alternatives. ## Target: Directus - **Type:** headless - **Official site:** https://directus.io - **Pricing:** https://directus.io/pricing - **Last reviewed:** 2026-07-28 ## Key advantages Directus instantly wins points with us because, well… it has a rabbit for a logo. But beyond that, it's genuinely a solid database-first CMS. If you like the idea of a CMS that sits directly on top of your SQL database without any abstractions, Directus feels incredibly natural. You can self-host it, tweak it endlessly, and treat your schema exactly the way you want. Editors get a no-code admin UI that feels polished, developers get REST and GraphQL out of the box, and teams get workflows, versioning, automations, and proper permission controls. It's flexible, fast, and a great fit if your content is really just structured data waiting for a smarter interface. If your project revolves around relational data and you want total control from database to API, Directus is one of the cleanest ways to build it. ### Intuitive non-technical interface Editors get a clean, no-code admin panel that feels more like a productivity tool than a CMS. Non-technical teams can handle content updates without pinging developers every five minutes. ### Database-agnostic SQL integration Directus plugs straight into your SQL database and turns it into an API instantly. If your data lives in Postgres, MySQL, MariaDB, or SQLite, it just works. ### Field-level permission controls You can lock down every field, table, and action with granular role-based permissions. Perfect for teams that care about security and don't want interns accidentally deleting production data. ### Built-in REST/GraphQL support Your entire database is instantly exposed as both REST and GraphQL endpoints. Just plug into your frontend and start shipping. ### Extensive UI customization Directus lets you tweak the interface, add custom views, tailor layouts, and build the exact editing experience your team needs. It feels like a CMS you can actually shape instead of fight. ### Real-time collaborative editing Multiple contributors can work in the Studio at the same time without stepping on each other's toes. Collaborative editing shipped in February 2026, so drafts stay clean and the whole workflow feels built for modern teams, not 2010 intranets. ### AI Assistant with image and PDF support The built-in AI Assistant can process images and PDFs as well as text, and it works across OpenAI, Anthropic, and Google Gemini. It's a genuinely useful addition rather than a feature-list checkbox. ## Key pain points Directus looks fantastic in demos, but things get rocky once you actually try to use it at scale. The Team cloud plan is $499/month (or $599 billed monthly) and caps you at 10 SSO seats, 50 collections, and 20 flows. Grow past those limits and you're straight into custom Enterprise territory. The v12 move to the Monospace Sustainable Core License (MSCL) has also rattled the community; smaller orgs get a free Innovation Grant, but larger teams are navigating a licensing landscape that changed significantly from what they signed up for. On the dev side, updates can introduce breaking changes, the documentation doesn't always keep pace, and the extension ecosystem is pretty thin. Localization is technically supported but fiddly and easy to misconfigure, and large datasets make the UI noticeably sluggish. And if you want anything deeply custom, you're suddenly living in Vue.js land, which is not where most teams want to spend their weekends. ### Breaking update changes Directus explicitly does not follow semver: any release may include breaking changes. The v10 to v11 upgrade hit schema changes and dropped fields, and the v11 UI scale change (px to rem) broke extensions hardcoding pixel values. Plan your upgrade windows carefully. ### Limited extension ecosystem The plugin ecosystem is still pretty bare. Anything even mildly niche ends up becoming a "let's just custom build it" moment, which defeats the purpose of picking a CMS with extensions. The marketplace launched in beta in early 2024 and is still maturing. ### Complex localization setup Yes, it supports multilingual content, but setting it up feels like assembling IKEA furniture blindfolded. It works, but expect extra config, extra steps, and extra patience. ### Version history still tier-gated Global draft versions now ship automatically with every versioned item (no manual setup required as of March 2026), which is a genuine improvement. Full version history, rollbacks, and controlled releases are still locked behind higher-tier plans, though, so if predictable publishing workflows are a must, check your tier carefully before committing. ### Vue.js knowledge requirement Custom interfaces and deeper tweaks need Vue.js, so if your team only speaks React, prepare for a small identity crisis (or a hiring plan). ### Large dataset performance issues Heavy tables and deeply relational data can slow down queries and the UI: community reports show 25K-row datasets where raw SQL runs in milliseconds but the Directus API takes 20+ seconds, largely due to internal query overhead and no auto-indexes on foreign keys. ## Frequently asked questions ### What makes migrating from KeystoneJS difficult? KeystoneJS stores data through Prisma, so the database layer is well-structured and easy to export. The harder part is replacing everything Keystone doesn't give you. Most Keystone projects have custom-built preview systems, publishing workflows, and access control logic that are tightly coupled to the Node.js backend. Rebuilding those features in a new CMS takes planning. We typically budget 4 to 8 weeks for a Keystone migration depending on how much custom infrastructure the team has built around it. ### Why do teams move away from KeystoneJS? Deployment complexity is the number one reason. Teams love Keystone during local development, then hit a wall getting it reliably into production. The Docker images can balloon past a gigabyte, the docs don't cover production hosting well, and there's no managed hosting option. The small community compounds this problem. When you hit an edge case, there are fewer people who've solved it before. Content editors also struggle with the admin UI, which lacks visual editing, live preview, and built-in publishing workflows that competing platforms ship by default. ### How do we extract our content from KeystoneJS? Since Keystone uses Prisma ORM, your content lives in standard PostgreSQL, MySQL, or SQLite tables with clean schemas. You can export directly from the database using SQL dumps or Prisma's query API. The content model is defined in your TypeScript codebase, so mapping fields to a new CMS is straightforward. We write automated scripts that handle the data transformation, including resolving relationships between lists and migrating file references. For a project with 20 to 50 Keystone lists, extraction and transformation usually takes 1 to 2 weeks.