Code Fundi Mcp

io.github.FelixWaweruPublished by GitHub account FelixWaweruv0.1.4Updated Oct 24, 2026

CodeFundi MCP server for codebase mapping and blast-radius analysis.

Overview

AI-generated overview

A local MCP server that connects an AI assistant to the Code-Fundi API for searching, indexing and mapping code repositories and analyzing the impact of changes

What you can do
Search and research indexed codebases with semantic and grep search, plus AI-synthesized analysis of matching code. List, index and check the status of repositories, and browse a public catalog of indexed repos. Map cross-repository dependencies, get repository blueprints and run blast-radius analysis of what a change breaks. List repository files and fetch AI-generated file documentation. Review query history, conversations, usage, activity and language statistics. Chat with Fundi AI, list chat models and model limits, and manage API keys through OTP or password sign-in.
When to use it
Useful when an assistant works on codebases that are indexed in Code-Fundi and you want natural-language search, repository overviews, or impact analysis before merging a change. Also relevant for exploring cross-repository dependencies and public repositories.
Requirements
Runs locally as a Node.js process, typically via npx or a global npm install of the package. Needs network access to the Code-Fundi API. Authentication is either a CODEFUNDI_API_KEY environment variable or an in-session OTP or password sign-in; CODEFUNDI_BASE_URL can override the API endpoint. A free-tier account is created on first signup.
Before you install
The server handles account credentials: OTP codes and passwords are sent to the Code-Fundi API, and the password goes in the X-CodeFundi-Auth-Password header. The API key is held in memory only and is lost on restart. Tools can index repositories, regenerate or disable API keys, and send code queries to a third-party service. Blast-radius analysis is listed as PRO+ tier.

Installation

In SourceWeft

  1. Open Code Fundi Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Code Fundi

The Codebase Map for AI Agents, Teams and Applications.

[Code Fundi]

[Image] [Image]
[Image] [Image] [Image]

Code-Fundi MCP Server

A production-grade Model Context Protocol (MCP) server for the Code-Fundi API: a codebase map and blast-radius engine for AI coding agents. Enables any MCP-compatible AI assistant (Claude, Claude Code, Cursor, Windsurf, VS Code Copilot, and more) to search, research, index, and scan code repositories through natural language, and see what a change breaks before it ships.

Built with FastMCP (TypeScript) and Zod for schema validation.

Getting started for AI agents

If you are an AI agent (or configuring one) and no CODEFUNDI_API_KEY is set, you must self-authenticate before using search, repos, chat, or other authenticated tools. The auth tools (code-fundi-auth-*) work without a pre-existing key.

Which path to use

SituationAction
CODEFUNDI_API_KEY is set in MCP configUse any tool immediately — the server sends it as X-API-Key on every request
No API key configuredRun the OTP self-auth flow below (recommended) or password sign-in

Zero-config MCP setup (no API key)

You can start the MCP server without CODEFUNDI_API_KEY and authenticate at runtime:

json
{  "mcpServers": {    "code-fundi": {      "command": "npx",      "args": ["-y", "@codefundi/code-fundi-mcp"]    }  }}

Self-authenticate with OTP (recommended)

Code-Fundi uses Supabase-backed auth (POST /v2/auth/authenticate, /v2/auth/verify, /v2/auth/resend). OTP emails contain a 6-digit code — magic links are not supported on this path.

  1. Ask the human user for their email address.
  2. Call code-fundi-auth-authenticate with:
    • auth_mode: "otp"
    • email: the user's email
    • should_create_user: true for a new account, false for a returning user
  3. Tell the user to check their inbox for a 6-digit code. The API may return verification_required: true and api_key.key_state: "agent_pending" until verification completes.
  4. Ask the user for the 6-digit OTP (human-in-the-loop — you cannot guess or bypass this step).
  5. Call code-fundi-auth-verify with the same email and the token (6 digits).
  6. On success, the MCP server automatically configures the API key in memory for all subsequent tool calls in this session.

If the code expired or was not received, call code-fundi-auth-resend with the same email, then repeat step 5.

Example dialogue:

Agent: What email should I use to sign in to Code-Fundi?User:  [email protected]Agent: [calls code-fundi-auth-authenticate] I've sent a 6-digit code to [email protected]. Please paste it here.User:  482913Agent: [calls code-fundi-auth-verify] You're signed in. I can now search and index your repositories.

Password sign-in (alternative)

For existing accounts with a password, call code-fundi-auth-authenticate with auth_mode: "password", the user's email, should_create_user: false, and the password parameter. The MCP client sends the password only in the X-CodeFundi-Auth-Password header (never in the JSON body). Production requires HTTPS. Returning users may receive an active API key immediately without a separate verify step.

After authentication

  • The API key is held in memory for the lifetime of the MCP server process. It is not persisted across IDE or MCP restarts.
  • Recommend the user add the key to their MCP config as CODEFUNDI_API_KEY so future sessions start authenticated.
  • A FREE-tier account and API key are created automatically on first signup.

Errors

HTTP statusWhat to do
401 UnauthorizedNo valid key — run the OTP flow above, or set CODEFUNDI_API_KEY
429 Too many requestsAuth endpoints are rate-limited per IP; wait for Retry-After seconds, then retry

Features

Every tool below is backed by the same codebase map: structural dependencies, call graph, and blast radius, indexed once and queried in milliseconds.

  • 🔍 Semantic & grep code search across your indexed codebase map
  • 🧠 AI-powered research: search plus AI analysis in a single call
  • 📦 Repository management: index, status, README, listing, public catalog
  • 🛰️ Repository intelligence: cross-repo dependency map, blueprint, Blast-Radius Guard (impact analysis before you merge)
  • 📄 File documentation: AI-generated docs for any indexed file
  • 📊 Usage statistics: query usage, activity, language breakdowns
  • 🔐 Agent-driven authentication: sign up/sign in via OTP without pre-configured keys
  • 💬 AI chat & model insight: direct conversation with Code-Fundi AI, model catalog, and per-tier limits

Quick Start

Install from npm (recommended)

Install the package (includes a pre-built dist/). The code-fundi-mcp binary is on your PATH when installed globally, or available via npx without cloning the repo:

bash
npm install -g @codefundi/code-fundi-mcp

Or add it to a project:

bash
npm install @codefundi/code-fundi-mcp

Install from source (this repository)

bash
git clone https://github.com/Code-Fundi/code-fundi-mcp.gitcd code-fundi-mcpnpm installnpm run build

Configure

Option A — API key (fastest): set your key as an environment variable:

bash
export CODEFUNDI_API_KEY=your_api_key_here

Option B — no API key: skip the env var and let the agent self-authenticate at runtime. See Getting started for AI agents.

Zero-config MCP example (no env block):

json
{  "mcpServers": {    "code-fundi": {      "command": "npx",      "args": ["-y", "@codefundi/code-fundi-mcp"]    }  }}

Use with Claude Desktop

After a global install (npm i -g @codefundi/code-fundi-mcp), point MCP at the published binary (no path to dist/index.js required):

json
{  "mcpServers": {    "code-fundi": {      "command": "@codefundi/code-fundi-mcp",      "env": {        "CODEFUNDI_API_KEY": "your_api_key_here"      }    }  }}

If the binary is not on your PATH, use npx (downloads or uses the local package and runs the same entrypoint):

json
{  "mcpServers": {    "code-fundi": {      "command": "npx",      "args": ["-y", "@codefundi/code-fundi-mcp"],      "env": {        "CODEFUNDI_API_KEY": "your_api_key_here"      }    }  }}

Use with Cursor

Same pattern as Claude: command plus optional args only, no manual path to the repo:

json
{  "mcpServers": {    "code-fundi": {      "command": "npx",      "args": ["-y", "@codefundi/code-fundi-mcp"],      "env": {        "CODEFUNDI_API_KEY": "your_api_key_here"      }    }  }}

When developing inside this repo, you can run npm run dev or npx tsx src/index.ts without building first.

Development Mode

bash
npm run dev                          # Run with tsx (no build needed)npx fastmcp inspect src/index.ts    # Open MCP Inspector UInpx fastmcp dev src/index.ts        # Test with MCP CLI

Tools Reference (27 tools)

Covers the Code-Fundi V2 API for codebase mapping and blast-radius analysis: search (including search-with-chat / research), repositories (list, index, status, readme, public catalog), repository intelligence (map, blueprint, radius), files, history, statistics, API keys, authentication, plus Fundi chat (POST /v1/fundi/chat) and the V2 model catalog / limits (GET /v2/models, GET /v2/models/limits).

Search

ToolDescription
code-fundi-searchSemantic and grep search across your indexed codebase map, with filters
code-fundi-researchSearch plus AI-synthesized analysis of matching code

Repositories

ToolDescription
code-fundi-list-reposList indexed repositories with pagination
code-fundi-index-repoIndex a new GitHub repository into your codebase map
code-fundi-repo-statusCheck repository indexing status
code-fundi-repo-readmeGet repository README documentation (deprecated, prefer blueprint)
code-fundi-list-public-reposBrowse the global catalog of indexed public repositories (no key needed)

Repository Intelligence

ToolDescription
code-fundi-repo-mapCross-repository dependency map: how services and packages actually connect
code-fundi-repo-blueprintREADME plus dependency and convention overview (successor to repo-readme)
code-fundi-repo-radiusBlast-Radius Guard: every file and function that breaks before you merge (PRO+)

Files

ToolDescription
code-fundi-list-filesList files in a repository
code-fundi-file-docsGet AI-generated file documentation

History

ToolDescription
code-fundi-list-historyList query history with filters
code-fundi-history-itemGet full details of a history entry
code-fundi-conversationGet conversation thread messages

Statistics

ToolDescription
code-fundi-usage-statsQuery-type usage breakdown
code-fundi-activity-statsDaily activity statistics
code-fundi-language-statsProgramming language usage

Authentication

ToolDescription
code-fundi-auth-authenticateStart OTP/password auth flow (no key needed)
code-fundi-auth-verifyVerify OTP code and obtain API key
code-fundi-auth-resendResend OTP verification email
code-fundi-list-api-keysList API keys (masked)
code-fundi-regenerate-api-keyRegenerate API key
code-fundi-disable-api-keyDisable an API key by ID (DELETE /v2/keys/{key_id})

Chat & Models

ToolDescription
code-fundi-chatFundi AI chat (POST /v1/fundi/chat; streamed responses are collected to text)
code-fundi-list-modelsList the curated chat model catalog (GET /v2/models)
code-fundi-model-limitsGet AI model limits and tier configuration (GET /v2/models/limits)

Authentication

Two modes: pre-configured API key (CODEFUNDI_API_KEY in MCP config) or agent-driven OTP/password auth at runtime. Full step-by-step instructions, tool names, and error handling are in Getting started for AI agents at the top of this README.

Environment Variables

VariableRequiredDefaultDescription
CODEFUNDI_API_KEYNo*—API key for authentication
CODEFUNDI_BASE_URLNohttps://api.codefundi.appAPI base URL override

* Required unless using agent-driven auth tools.

License

MIT

Source: README.md at commit ad33034

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.4LatestSep 16, 2026

Score

40% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.