FeathersMCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@FeathersMCP How do I create a custom hook in Feathers v6?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
FeathersMCP
An MCP server that connects your AI assistant to the FeathersJS v6 documentation. Get accurate, source-linked answers about hooks, services, authentication, and more, directly in your IDE.
Why use FeathersMCP?
AI assistants answering from memory often mix up Feathers versions, invent APIs, or miss edge cases. FeathersMCP gives your assistant live access to the real v6 docs so answers come with a source_url you can verify.
Without FeathersMCP | With FeathersMCP |
Training-data guesses, risk of outdated APIs | Searches and returns actual v6 documentation pages |
No citations | Every answer includes a link to |
No community patterns | Optional community knowledge base (real-world tutorials) |
Related MCP server: Whoosh RAG MCP
Prerequisites
Requirement | Link |
Node.js (v18 LTS or later) | |
npm (bundled with Node) | |
An MCP-compatible AI client (see Installation) | (n/a) |
Supported clients: VS Code (with GitHub Copilot), Cursor, Claude Desktop, Zed, and any other MCP host.
Each client uses its own MCP config location. Follow the section for your client below (Step 1 shows the exact file or setting to edit).
Installation
The server command is the same everywhere; only the config file location and JSON shape depend on your MCP host. Open the section for the client you use.
VS Code (GitHub Copilot)
Step 1. Create .vscode/mcp.json inside your project folder:
{
"servers": {
"feathersjs": {
"type": "stdio",
"command": "npx",
"args": ["feathersjs-mcp@latest"]
}
}
}Step 2. Reload the window — open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and run Developer: Reload Window.
Step 3. Open the Command Palette again, run MCP: List Servers, and confirm feathersjs appears in the list.
If it shows as Stopped, click it and select Start Server.
If VS Code prompts you to trust the server, click Allow. Without this the server will not run.
If it is listed but disabled, enable it from the same panel.
Step 4. Open GitHub Copilot Chat, switch to Agent mode, and ask:
What FeathersJS topics are covered in the official docs?
If the server is connected, the assistant should call a tool like get-menu or search-doc. If no tool call appears, check that the server is running in MCP: List Servers and that you are in Agent mode, not Ask or Edit mode.
Cursor
Step 1. Create the config file. Two options:
Project only:
.cursor/mcp.jsoninside your project folder (can be committed to git and shared with teammates)All projects:
~/.cursor/mcp.jsonin your home directory (local to your machine only)
If both files exist, the project-level file takes priority.
{
"mcpServers": {
"feathersjs": {
"command": "npx",
"args": ["feathersjs-mcp@latest"]
}
}
}Step 2. Reload the window — open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and run Developer: Reload Window.
Step 3. Open Cursor Settings → Tools & MCP and confirm feathersjs appears with a green status indicator. If the toggle next to it is off, turn it on. Adding the config file alone is not enough — the server must be enabled in this panel.
Step 4. Open Cursor Chat in Agent mode and ask:
How do Feathers hooks work in v6?
If the server is active, the assistant should call search-doc and return a result with a source URL. If no tool is called, check that the server is toggled on in Settings → Tools & MCP and that you are in Agent mode, not Ask or Edit mode.
Claude Desktop
Step 1. Open the config file for your OS:
macOS:
~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\Claude\claude_desktop_config.json
Step 2. Add the server entry (create the file if it does not exist):
{
"mcpServers": {
"feathersjs": {
"command": "npx",
"args": ["-y", "feathersjs-mcp@latest"]
}
}
}Step 3. Fully restart Claude Desktop — quit and reopen, do not just reload. The first run downloads the package automatically.
Step 4. Ask a test question:
What authentication strategies does Feathers v6 support?
If no tool is called, check Claude's MCP logs (Help → Open Logs Folder) for connection errors. See Claude Desktop MCP docs for more detail.
Zed
Step 1. Add the server to your Zed settings. Open Zed Settings (Cmd+,) and add a context_servers block:
{
"context_servers": {
"feathersjs": {
"command": {
"path": "npx",
"args": ["-y", "feathersjs-mcp@latest"]
}
}
}
}Alternatively, open the Agent Panel (Cmd+?), click the settings icon, and use Add Custom Server to register the same command.
Step 2. Check the status indicator next to the server name in the Agent Panel. A green dot means it is active. If it is grey or shows an error, verify npx is on your PATH by opening Zed from a terminal: zed .
Step 3. Note that Zed may ask you to approve tool calls before they run — this is expected. Confirm when prompted.
Step 4. Ask a test question in the Agent Panel:
What topics are covered in the official FeathersJS v6 docs?
If the server is connected, it should call get-menu or search-doc. Mentioning feathersjs by name in your prompt can help Zed route to the right server.
Other MCP hosts
Register the same command — npx with args ["-y", "feathersjs-mcp@latest"] — using your client's MCP config format. The MCP clients list links setup docs for each supported host.
Optional: pin the version locally
npm install feathersjs-mcpLocks the package in package.json so all contributors use the same release. Swap "npx" for "node" and point args at node_modules/feathersjs-mcp/build/index.js.
Agent skill (optional)
The skill file adds a workflow guide so your assistant searches docs and community before answering from memory.
# Install
npx feathersjs-mcp@latest install-skill
# Uninstall
rm -rf .agents/skills/feathersjs-mcpWorks with Cursor, Claude Code, Windsurf, Gemini CLI, and any tool that follows the agentskills.io standard. Re-run the install command to update it.
Prompts to try
These natural-language questions demonstrate what FeathersMCP is good at. Paste any of them into your AI chat after installation.
What you ask | Tool triggered |
|
|
|
|
|
|
|
|
You don't need to mention tool names. The server picks the right one based on your question.
Troubleshooting
Server not connecting / tools never appear Check the client-specific step for your MCP host above — most clients require you to manually enable or start the server after adding the config file. Also verify the config file is in the correct location for your client (see the table in Prerequisites).
macOS: spawn npx ENOENT
Open your project from a terminal (cursor . / code . / zed .) so the IDE inherits your shell PATH. Or install locally and use the full path to node (from which node):
{
"mcpServers": {
"feathersjs": {
"command": "/opt/homebrew/bin/node",
"args": ["node_modules/feathersjs-mcp/build/index.js"]
}
}
}Tools appear but no tool is called in chat Make sure you are in Agent mode — tools are not available in Ask or Edit mode. In Cursor, also verify the server toggle is on in Settings → Tools & MCP.
Available tools
Tool | Description |
| Full doc index (categories, paths, IDs) |
| Full-text search across 47 official v6 pages |
| Full page content and code examples (by id, path, or title) |
| Schema of the local docs database |
| Search community tutorials and patterns |
| Full community post by id |
| Generate a pre-filled GitHub issue to submit a community post |
Links
npm package: https://www.npmjs.com/package/feathersjs-mcp
GitHub repo: https://github.com/daffl/FeathersMCP
FeathersJS v6 docs: https://feathersjs.com / https://v6.feathersjs.com
MCP standard: https://modelcontextprotocol.io
Agent skills: https://agentskills.io
Repository documentation (optional — for people working in the repo)
If you only use FeathersMCP from npm (for example npx feathersjs-mcp in your MCP config), you do not need anything under docs/—the sections above are enough.
The docs/ folder supports transfer of ownership, onboarding new maintainers, and ongoing operations. Layout: docs/project/ (ownership transfer, setup, architecture) and docs/operations/ (CI/CD, releases, D1 — DevOps-style docs for maintainers).
Document | Who it is for |
Incoming maintainers — scope, limitations, operational ownership | |
Developers — clone, build, and verify from a clean machine | |
Developers — architecture and data flow | |
Maintainers — GitHub Actions | |
Maintainers — npm and skill release | |
Maintainers — D1 reset and schema |
Available Tools
7 toolsget-community-postA
Fetch the full content of a community contribution by id (from search-community results).
Inputs: id (required).
get-docA
Fetch a full official FeathersJS doc page: returns complete page text plus all code examples. Look up by id, path, or title. Use after get-menu (paths/ids), after search-doc, or when the user gave an exact id/path/title. Prefer id or path over title.
Inputs: id, path, title.
get-schemaA
Get the database schema for FeathersJS documentation tables
Inputs: no named parameters.
search-communityA
Search real-world FeathersJS solutions from the community.
PRIMARY use cases:
Implementation, integrations, and architecture patterns
Debugging issues or unexpected behavior
Workarounds, fixes, and best practices not in docs
You MUST use this when:
The user asks how to implement or integrate something
The user reports something not working or asks why
The solution is not clearly in a single doc page
After calling:
Use
get-community-postif a result is relevantInclude at least one community insight in your answer
Returns: results with id for get-community-post.
Inputs: query (required).
search-docB
Search the FeathersJS documentation for relevant content. You can make multiple calls to this tool to refine your search. Call get-schema and get-menu first to craft better queries.
Inputs: query (required), limit, category, query_type.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool has a distinct role: get-menu and get-schema provide structural context, search-doc and search-community handle discovery, get-doc and get-community-post retrieve full resources, and share-knowledge handles contribution submission. The doc vs. community split is consistent and clearly reflected in descriptions, so an agent can easily choose the right tool.
All tool names follow the same lowercase hyphenated verb-noun convention: get-doc, get-community-post, get-menu, get-schema, search-doc, search-community, share-knowledge. There are no mixed casing styles or inconsistent verb patterns.
Seven tools is well scoped for a documentation and community knowledge server. The set covers navigation, schema inspection, documentation search/retrieval, community search/retrieval, and contribution sharing without obvious redundancy.
The server covers the main lifecycle: orient with menu/schema, search docs and community, retrieve full items, and share knowledge. Minor gaps exist, such as no explicit listing or filtering of community posts by tag/author, but search largely compensates for those omissions.
Maintenance
Related MCP Connectors
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
An MCP server that gives your AI access to the source code and docs of all public github repos
Official Duck Framework MCP server for AI-assisted development and project scaffolding.
An MCP server that integrates with Discord to provide AI-powered features.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceAn MCP server that enables AI assistants to access up-to-date documentation for Python libraries like LangChain, LlamaIndex, and OpenAI through dynamic fetching from official sources.1MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that provides full-text search over documentation using Whoosh, enabling AI assistants to find up-to-date, authoritative answers.MIT
- AlicenseNot gradedqualityBmaintenanceAn MCP server that provides tools for retrieving and processing documentation through vector search, enabling AI assistants to augment their responses with relevant documentation context.18 npmMIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that gives AI coding assistants access to up-to-date API documentation via RAG by crawling documentation sites, indexing them into a vector store, and enabling semantic queries.MIT