DevKit for Strapi MCP Server
Provides AI agents with ground-truth knowledge of Strapi projects by reading real schema files, enabling accurate queries and safe refactoring operations.
README
<p align="center"> <img src="assets/icon.svg" width="120" alt="DevKit for Strapi logo" /> </p>
<h1 align="center">DevKit for Strapi</h1>
<p align="center"> Accurate, project-aware tooling for Strapi <em>magic strings</em> — for your editor <strong>and</strong> your AI agent.<br/> Where Copilot <strong>guesses</strong>, DevKit for Strapi <strong>knows</strong>: it reads your real <code>schema.json</code>. </p>
Why
Strapi runs on magic strings — UIDs like api::article.article, service/controller refs, policy and
middleware names, component UIDs, route handlers. In most editors they have no autocomplete, no
validation, and broken navigation, and a typo fails silently at runtime. AI coding assistants make
it worse: they happily hallucinate UIDs and method names that don't exist in your project.
DevKit for Strapi reads the real content-types, components, services, controllers, policies, middlewares and routes of every Strapi project in your workspace, and turns those strings into first-class, verified code — in two surfaces that share one engine:
- a VS Code extension for humans, and
- an MCP server that gives AI agents the same ground truth.
The rule everywhere: guarantee, don't guess. When something can't be proven (a non-literal string, an unverifiable external plugin), DevKit stays silent rather than show a false positive.
🧩 The VS Code extension
Inside a Strapi project — zero config, no type generation:
- Autocomplete — inside a magic string, suggests the real values for the context at your cursor: UIDs, services, controllers, policies, middlewares, components, and a content-type's route actions.
- Diagnostics — underlines references that don't exist (with quick fixes that suggest the closest
match), and flags Strapi v4 patterns (
entityService,data.attributes) used in a v5 project. - Go to definition — <kbd>Ctrl/Cmd</kbd>-click a string to jump to the right target: a UID → its
schema.json, a service/controller/policy/middleware → its file, a routehandler→ the controller method, a component → its JSON, aplugin('a').service('b')chain → the plugin's service. - Find references & CodeLens — a “N references” lens on every definition, including per-method
usages (
strapi.service('x').find()calls and route handlers) the TypeScript server can't see, becausestrapi.service(...)is typedany. A second “N incoming relations” lens onschema.json. - Rename (<kbd>F2</kbd>) — on an entity or a method: propagates to every call-site and route
handler, renames the underlying file/folder, and updates relation
targets. No grep-and-pray. - Hover — describes what a string resolves to, with the signature for service methods and usage insights (incoming relations, route handlers, data usages).
🤖 The MCP server — ground truth for AI agents
The same engine, exposed over the Model Context Protocol so coding agents (Claude Code, Copilot agent mode, Cursor, …) query your project's real values instead of inventing them. 29 stdio tools, in three layers:
Know — read the project's truth
list_projects · list_content_types · list_components · list_artifacts · get_schema · resolve ·
validate_reference · find_references · list_routes
Understand — health & impact analysis
list_unused (dead code) · list_broken_refs · coverage · find_relation_usages · list_refs (glob) ·
dependencies / dependents (the dependency graph, for cut analysis)
Refactor — plan → review → apply
plan_rename_method / plan_rename_entity · plan_move / plan_move_entities · plan_change_relation ·
plan_rename_attribute · create_plugin · extract_to_plugin, then apply_edits / apply_rename.
Every refactor returns a contractual, reviewable plan (text edits + file creates/renames/deletes + content fingerprints). Applying is explicit and safe by construction: the executor is confined to your discovered project root(s) (symlink-resolved — no escape), refuses to silently overwrite, verifies the fingerprints first (a stale plan is rejected, not applied), is best-effort transactional (rolls back on a mid-write failure), and self-verifies afterward (re-checks for broken refs). Garantir, ne pas deviner — applied to writes too.
Set it up
- VS Code, Cursor, Windsurf, Antigravity — the server is bundled in the extension and auto-registered with the editor's MCP support. Nothing to configure: install the extension and your agent can use the tools.
- Other MCP clients — run
devkit-for-strapi-mcpas a stdio server, passing your project path. It also accepts the workspace folder via the client'srootscapability (so it just works wherever you open it), and a tool (add_project) to register a project on demand.
For Claude Code, Claude Desktop, or any standalone MCP client:
{
"mcpServers": {
"devkit-for-strapi": {
"command": "npx",
"args": ["-y", "devkit-for-strapi-mcp", "/absolute/path/to/your/strapi-project"]
}
}
}
Try it
Once it's wired, ask your agent:
- "Which content types exist in this project?"
- "What uses
api::article.article?" - "Is
api::article.artclea valid reference?"
It answers from your real schema.json — not from a guess.
Accurate by design
- Reads your real schema — answers come from your project's files, not from a model's guess.
- Multi-project — discovers every Strapi project in the workspace by content (scanning for
@strapi/strapi), never assuming workspace root = Strapi root. References resolve against the project that owns the edited file; an ambiguous request returns candidates, never a silent choice. - Strapi v4 and v5 — the version is detected per project; v4-in-v5 patterns are flagged. Knows the
difference between a
collectionTypeand asingleType(their auto-CRUD actions differ) and a schema-only content-type's auto-generated controller. - JavaScript & TypeScript — identical behavior for ESM and CommonJS projects (challenged by a JS⇄TS parity test).
- Runs everywhere — VS Code and its forks (Cursor, Windsurf, Google Antigravity) via Open VSX and the VS Code Marketplace.
Install
Editor — search “DevKit for Strapi” in the Extensions view (VS Code Marketplace), or install from Open VSX for Cursor / Windsurf / VSCodium. Open a Strapi project and it works.
Requirements — a Strapi v4 or v5 project, and VS Code 1.101+ (or a compatible fork). No
configuration, no type generation, no extra setup.
The MCP server needs no separate install in those editors — it ships inside the extension.
Architecture
A pnpm monorepo around one pure engine:
| Package | Role |
|---|---|
devkit-for-strapi-core |
The editor-agnostic engine — all Strapi knowledge, reads IO through a FileSystem seam, pure TypeScript. |
devkit-for-strapi |
The VS Code extension (providers, watcher) — glue, no Strapi logic. |
devkit-for-strapi-mcp |
The stdio MCP server — a third client of the same engine. |
The same core powers the editor, the MCP server, and the tests — so a fact verified once is true on every surface.
Trademark
Independent, community-built tool — not affiliated with, endorsed by, or sponsored by Strapi SAS. “Strapi” is a trademark of Strapi SAS, used here only to describe compatibility.
License
The core engine and the free editor/MCP wedge are free to use (licensed MIT)
(devkit-for-strapi-core, plus the read/navigation/diagnostics surface of the extension and the
read-only MCP tools).
The Pro write/refactor engine (propagated rename, move/extract, schema edits, plan→apply) is a
separate package that is not included in this repository — the packages/pro here is an
MIT-licensed stub that only preserves the public API so the free tier builds and runs. The real Pro
engine is licensed under PolyForm Shield 1.0.0 and is bundled into the published VS Code
extension and MCP server, so those distributed artifacts are governed by MIT AND PolyForm-Shield-1.0.0. Building from this repository produces the free tier; Pro features return a
"Pro required" upsell. Get Pro: https://devkit-for-strapi.paulrichez.fr/pro/.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.