jev-docs-mcp
by mohit716
README.md
# jev-docs-mcp
An [MCP](https://modelcontextprotocol.io) server that lets AI coding agents (Cursor, Claude Code, Codex, and other MCP clients) search and read the official [TypeSafe Jev documentation](https://docs.typesafe.ai).
It loads the docs from `docs.typesafe.ai/llms.txt` and `docs.typesafe.ai/llms-full.txt` and caches them in memory for an hour. No API key is needed.
## Tools
| Tool | What it does |
| --- | --- |
| `jev_docs_list` | Lists every docs page with its path and a one-line summary. |
| `jev_docs_search` | Full-text search across all pages; returns the best matches with snippets. |
| `jev_docs_read` | Returns a full page by path (`primitives/noul`), URL, or title. Long pages are paginated with `offset`. |
## Setup
Requires Node.js 18 or newer. Nothing to install: `npx` downloads and runs the package.
### Cursor
Add this to `~/.cursor/mcp.json` (or `.cursor/mcp.json` in a project):
```json
{
"mcpServers": {
"jev-docs": {
"command": "npx",
"args": ["-y", "jev-docs-mcp"]
}
}
}
```
Then open **Cursor Settings → Tools & MCP** and make sure `jev-docs` is enabled.
### Claude Code
```bash
claude mcp add jev-docs -- npx -y jev-docs-mcp
```
### Any other MCP client
Run `npx -y jev-docs-mcp` as a stdio server.
## Development
```bash
git clone https://github.com/mohit716/jev-docs-mcp.git
cd jev-docs-mcp
npm install
npm test
```
This starts the server, calls each tool against the live docs, and prints the results.
## Not affiliated with TypeSafe
This is an independent project. The documentation content belongs to TypeSafe AI.
TDQS
A4.4/5.0
Scored across 3 tools
Disambiguation5/5
Each tool has a clearly distinct purpose: list enumerates all pages, search performs full-text lookup, and read fetches full page content. The descriptions explicitly clarify how they differ, leaving no realistic ambiguity.
Naming Consistency5/5
All three tools follow the same predictable pattern: jev_docs_ prefix plus a consistent verb (list, search, read). There is no mixing of conventions.
Tool Count5/5
Three tools is a well-scoped minimal set for a documentation server: discover, search, and read. Each tool earns its place without redundancy.
Completeness5/5
The surface covers the full read-only documentation lifecycle: discovery via list, targeted lookup via search, and full content retrieval via read with pagination support. No obvious gaps remain.