1- # CodeIntel MCP
1+ <p align =" center " >
2+ <!-- Logo placeholder - replace with actual logo -->
3+ <h1 align =" center " >OpenCodeIntel</h1 >
4+ </p >
5+
6+ <p align =" center " >
7+ <strong >AI-powered semantic code search for your repositories</strong >
8+ </p >
9+
10+ <p align =" center " >
11+ <a href =" https://github.com/OpenCodeIntel/opencodeintel/actions/workflows/ci.yml " >
12+ <img src="https://github.com/OpenCodeIntel/opencodeintel/actions/workflows/ci.yml/badge.svg" alt="CI Status" />
13+ </a >
14+ <a href =" https://github.com/OpenCodeIntel/opencodeintel/blob/main/LICENSE " >
15+ <img src="https://img.shields.io/github/license/OpenCodeIntel/opencodeintel" alt="License" />
16+ </a >
17+ <a href =" https://github.com/OpenCodeIntel/opencodeintel/releases " >
18+ <img src="https://img.shields.io/github/v/release/OpenCodeIntel/opencodeintel?include_prereleases" alt="Release" />
19+ </a >
20+ </p >
21+
22+ <p align =" center " >
23+ <a href =" #quick-start " >Quick Start</a > •
24+ <a href =" ./docs/deployment.md " >Deployment</a > •
25+ <a href =" ./docs/mcp-setup.md " >MCP Integration</a > •
26+ <a href =" ./CONTRIBUTING.md " >Contributing</a >
27+ </p >
228
3- ** MCP server for AI-powered codebase intelligence.** Semantic search, dependency analysis, and impact prediction for your repositories.
4-
5- ## The Problem
6-
7- AI coding assistants are powerful, but they're flying blind in large codebases:
8- - Can't semantically search across thousands of files
9- - Don't understand dependency relationships
10- - Can't predict what breaks when you change a file
11- - Have no context on team coding patterns
12-
13- ## The Solution
14-
15- CodeIntel is an MCP (Model Context Protocol) server that gives AI agents deep codebase understanding:
16-
17- ``` typescript
18- // Ask Claude (via MCP):
19- " Find authentication middleware in this repo"
20-
21- // CodeIntel semantically searches 10,000+ functions
22- // Returns exact implementations, not keyword matches
23- ```
24-
25- ** Built for production. Not a demo.**
26-
27- ## Key Features
28-
29- ### 🔍 Semantic Code Search
30- Search by meaning, not keywords. Find ` "error handling logic" ` even if functions are named ` processFailure() ` .
31-
32- ### 📊 Dependency Analysis
33- Visualize your entire codebase architecture. See which files are critical, which are isolated, and how everything connects.
29+ ---
3430
35- ### ⚡ Impact Prediction
36- Before changing a file, know exactly what breaks:
37- ```
38- src/auth/middleware.py
39- └─ 15 files affected (HIGH RISK)
40- ├─ src/api/routes.py
41- ├─ src/services/user.py
42- └─ ... + 12 more
43- ```
31+ <!-- Demo screenshot placeholder -->
32+ <!-- <p align="center">
33+ <img src="docs/assets/demo.png" alt="OpenCodeIntel Dashboard" width="800" />
34+ </p> -->
4435
45- ### 🎨 Code Style Analysis
46- Understand team patterns: naming conventions (camelCase vs snake_case), async adoption %, type hint usage.
36+ ## What is OpenCodeIntel?
4737
48- ### 🚀 Performance That Scales
38+ OpenCodeIntel gives AI coding assistants deep understanding of your codebase. It's an MCP server that provides semantic code search, dependency analysis, and impact prediction.
4939
50- ** Batch Processing:** 100x faster indexing
51- - Before: 40+ min for 1,000 functions (individual API calls)
52- - After: 22.9 sec (batch embedding requests)
40+ ** Search by meaning, not keywords.** Find "error handling logic" even when functions are named ` processFailure() ` .
5341
54- ** Incremental Indexing:** 700x faster re-indexing
55- - Full re-index: 51.4s
56- - Incremental (git diff): 0.07s
57- - Perfect for active development
42+ ## Features
5843
59- ** Supabase Caching:** 5x search speedup
60- - Cold search: 800ms
61- - Cached: 150ms
44+ - ** Semantic Search** - Vector-based code search that understands intent
45+ - ** Dependency Graph** - Visualize how your codebase connects
46+ - ** Impact Analysis** - Know what breaks before you change a file
47+ - ** Code Style Analysis** - Understand team patterns and conventions
48+ - ** MCP Integration** - Works directly with Claude Desktop
6249
6350## Quick Start
6451
65- ### 🐳 Docker (Recommended)
66-
67- ** Fastest way to get started:**
52+ ### Using Docker (Recommended)
6853
6954``` bash
70- # 1. Clone repo
7155git clone https://github.com/OpenCodeIntel/opencodeintel.git
7256cd opencodeintel
7357
74- # 2. Configure environment
7558cp .env.example .env
76- # Edit .env with your API keys
59+ # Add your API keys to .env
7760
78- # 3. Start everything
7961docker compose up -d
80-
81- # Frontend: http://localhost:3000
82- # Backend: http://localhost:8000
83- # Docs: http://localhost:8000/docs
8462```
8563
86- ** Full guide:** [ ` DOCKER_QUICKSTART.md ` ] ( ./DOCKER_QUICKSTART.md )
87- ** Troubleshooting:** [ ` DOCKER_TROUBLESHOOTING.md ` ] ( ./DOCKER_TROUBLESHOOTING.md )
64+ - Frontend: http://localhost:3000
65+ - Backend: http://localhost:8000
66+ - API Docs: http://localhost:8000/docs
8867
89- ---
90-
91- ### 📦 Manual Setup
68+ ### Manual Setup
9269
93- ### Prerequisites
94- - Python 3.11+
95- - Node.js 20+
96- - OpenAI API key
97- - Pinecone account
98- - Supabase project
99-
100- ### 1. Clone & Setup Backend
70+ ** Requirements:** Python 3.11+, Node.js 20+
10171
10272``` bash
73+ # Backend
10374cd backend
10475python -m venv venv
105- source venv/bin/activate # Windows: venv\Scripts\activate
76+ source venv/bin/activate
10677pip install -r requirements.txt
107-
108- # Configure .env
10978cp .env.example .env
110- # Add your API keys to .env
111- ```
112-
113- ### 2. Run Backend
114-
115- ``` bash
11679python main.py
117- # Server runs on http://localhost:8000
118- ```
11980
120- ### 3. Setup Frontend
121-
122- ``` bash
81+ # Frontend (new terminal)
12382cd frontend
12483npm install
12584npm run dev
126- # UI at http://localhost:5173
127- ```
128-
129- ### 4. Add a Repository
130-
131- ``` bash
132- # Via API
133- curl -X POST http://localhost:8000/api/repos \
134- -H " Authorization: Bearer dev-secret-key" \
135- -H " Content-Type: application/json" \
136- -d ' {"name": "zustand", "git_url": "https://github.com/pmndrs/zustand"}'
137-
138- # Or use the web UI
13985```
14086
14187## MCP Integration
14288
143- CodeIntel works as an MCP server with Claude Desktop. ** [ 📚 Full MCP Setup Guide ] ( ./docs/MCP_SETUP.md ) **
89+ Connect OpenCodeIntel to Claude Desktop for AI-powered code assistance.
14490
145- ** Quick Setup: **
91+ Add to your Claude Desktop config:
14692
14793``` json
148- // Add to Claude Desktop config
14994{
15095 "mcpServers" : {
151- "codeintel " : {
96+ "opencodeintel " : {
15297 "command" : " python" ,
15398 "args" : [" /path/to/opencodeintel/mcp-server/server.py" ],
15499 "env" : {
@@ -160,142 +105,41 @@ CodeIntel works as an MCP server with Claude Desktop. **[📚 Full MCP Setup Gui
160105}
161106```
162107
163- ** Available MCP Tools:**
164- | Tool | Description |
165- | ------| -------------|
166- | ` search_code ` | Semantic code search - finds code by meaning |
167- | ` list_repositories ` | View all indexed repos |
168- | ` get_dependency_graph ` | Visualize architecture and file connections |
169- | ` analyze_code_style ` | Team conventions and patterns |
170- | ` analyze_impact ` | Know what breaks before you change it |
171- | ` get_repository_insights ` | High-level codebase overview |
172-
173- Now ask Claude: * "What's the authentication logic in the user service?"* and it searches your actual codebase.
108+ ** Available tools:** ` search_code ` , ` list_repositories ` , ` get_dependency_graph ` , ` analyze_code_style ` , ` analyze_impact ` , ` get_repository_insights `
174109
175- ** [ → Complete setup guide with troubleshooting ] ( ./docs/MCP_SETUP .md ) **
110+ See [ MCP Setup Guide ] ( ./docs/mcp-setup .md ) for detailed instructions.
176111
177112## Architecture
178113
179114```
180- ┌─────────────┐
181- │ Frontend │ React + TypeScript + Tailwind
182- │ (Vite app) │ Dependency graphs, search UI
183- └──────┬──────┘
184- │
185- ┌──────▼──────┐
186- │ FastAPI │ Python backend
187- │ Backend │ /api/search, /api/repos/{id}/dependencies
188- └──────┬──────┘
189- │
190- ├─────► Pinecone (vector search)
191- ├─────► OpenAI (embeddings)
192- ├─────► Supabase (persistence)
193- └─────► Redis (caching)
194- ```
195-
196- ** Tech Stack:**
197- - ** Backend:** FastAPI, tree-sitter (AST parsing), OpenAI embeddings
198- - ** Vector DB:** Pinecone for semantic search
199- - ** Database:** Supabase (PostgreSQL) for metadata + caching
200- - ** Cache:** Redis for 5x search speedup
201- - ** Frontend:** React, TypeScript, Tailwind CSS, shadcn/ui, ReactFlow
202-
203- ## Performance Benchmarks
204-
205- Real numbers from indexing the Zustand repository (1,174 functions):
206-
207- | Metric | Value |
208- | --------| -------|
209- | Full indexing | 29.5s (39.7 functions/sec) |
210- | Incremental re-index | 0.07s (700x faster) |
211- | Batch embedding | 22.9s for 1,174 functions |
212- | Search (cold) | 800ms |
213- | Search (cached) | 150ms |
214-
215- ## Use Cases
216-
217- ** For AI Agents (via MCP):**
218- - Semantic code search during pair programming
219- - Understanding unfamiliar codebases
220- - Finding implementation patterns
221- - Impact analysis before refactoring
222-
223- ** For Development Teams:**
224- - Onboarding new engineers (visualize architecture)
225- - Code review prep (see change blast radius)
226- - Tech debt identification (find highly coupled files)
227- - Pattern enforcement (analyze style consistency)
228-
229- ## What Makes This Different
230-
231- ** Most code search tools:** Keyword matching (grep, GitHub search)
232- ** CodeIntel:** Understands * meaning* - finds ` error handling ` even if the function is called ` processFailure() `
233-
234- ** Most dependency tools:** Static analysis only
235- ** CodeIntel:** Combines AST parsing + semantic understanding + impact prediction
236-
237- ** Most demos:** In-memory, doesn't scale
238- ** CodeIntel:** Production-grade with Supabase persistence, Redis caching, incremental indexing
239-
240- ## Deployment
241-
242- ### 🐳 Local Development (Docker)
243- ``` bash
244- # Start all services
245- make dev
246-
247- # Or using docker compose
248- docker compose -f docker-compose.dev.yml up -d
249-
250- # Services available at:
251- # - Backend: http://localhost:8000
252- # - Frontend: http://localhost:3000
253- # - API Docs: http://localhost:8000/docs
254- ```
255-
256- ### ☁️ Production Deployment
257-
258- ** Backend + Redis → Railway**
259- ``` bash
260- # Automated deployment
261- ./scripts/deploy-railway.sh
262-
263- # Or manually:
264- railway login
265- railway init
266- railway up
115+ Frontend (React + TypeScript)
116+ ↓
117+ Backend (FastAPI + Python)
118+ ↓
119+ ┌────┴────┬────────────┐
120+ Pinecone Supabase Redis
121+ (vectors) (database) (cache)
267122```
268123
269- ** Frontend → Vercel**
270- ``` bash
271- # Automated deployment
272- ./scripts/deploy-vercel.sh
124+ ** Stack:** FastAPI, React, TypeScript, Pinecone, Supabase, Redis, tree-sitter
273125
274- # Or manually:
275- cd frontend
276- vercel --prod
277- ```
126+ ## Documentation
278127
279- ** 📚 Full deployment guide:** See [ DEPLOYMENT.md] ( DEPLOYMENT.md ) for complete instructions, environment variables, and troubleshooting.
128+ | Guide | Description |
129+ | -------| -------------|
130+ | [ Docker Quickstart] ( ./docs/docker-quickstart.md ) | Get running in 5 minutes |
131+ | [ Deployment] ( ./docs/deployment.md ) | Production deployment guide |
132+ | [ MCP Setup] ( ./docs/mcp-setup.md ) | Claude Desktop integration |
133+ | [ Docker Troubleshooting] ( ./docs/docker-troubleshooting.md ) | Common issues and fixes |
280134
281135## Contributing
282136
283- Built in a focused 2-week sprint to demonstrate production-grade AI development tooling .
137+ We welcome contributions! See [ CONTRIBUTING.md ] ( ./CONTRIBUTING.md ) for guidelines .
284138
285- Contributions welcome! Areas for improvement:
286- - Support for more languages (currently: Python, JS/TS)
287- - Advanced graph algorithms (find circular dependencies, suggest refactorings)
288- - GitHub integration (PR impact analysis)
289- - Team analytics (who writes what patterns)
139+ ** Quick links:**
140+ - [ Open Issues] ( https://github.com/OpenCodeIntel/opencodeintel/issues )
141+ - [ Good First Issues] ( https://github.com/OpenCodeIntel/opencodeintel/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22 )
290142
291143## License
292144
293- MIT License - use it, fork it, build on it.
294-
295- ## Built With
296-
297- Commitment to shipping production-grade AI tools. Not a side project. Not a demo. Real infrastructure that scales.
298-
299- ---
300-
301- ** Questions?** Open an issue or reach out.
145+ MIT License - see [ LICENSE] ( ./LICENSE ) for details.
0 commit comments