cctx-mcp
An MCP server that provides structure-aware code analysis (symbol trees, dependencies, docs) to reduce AI agent token consumption by up to 99%, along with Git commit intelligence.
README
<p align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://img.shields.io/badge/CCTX%E2%80%93MCP-0F172A?style=for-the-badge&logo=python&logoColor=3B82F6&labelColor=1E293B"> <img src="https://img.shields.io/badge/CCTX%E2%80%93MCP-F1F5F9?style=for-the-badge&logo=python&logoColor=2563EB&labelColor=FFFFFF" alt="CCTX-MCP"> </picture> </p>
<p align="center"> <strong>Cut AI agent token usage by 87%.</strong><br> Structure-aware code analysis via the <a href="https://modelcontextprotocol.io">Model Context Protocol</a> — returns symbol trees, dependencies, and docs instead of raw file contents. </p>
<p align="center"> <a href="https://pypi.org/project/cctx-mcp/"><img src="https://img.shields.io/pypi/v/cctx-mcp?style=flat-square&logo=pypi&logoColor=white&label=PyPI&labelColor=1E293B&color=3B82F6" alt="PyPI"></a> <a href="https://pypi.org/project/cctx-mcp/"><img src="https://img.shields.io/pypi/pyversions/cctx-mcp?style=flat-square&logo=python&logoColor=white&label=Python&labelColor=1E293B&color=3B82F6" alt="Python"></a> <a href="https://github.com/nikondrat/cctx-mcp"><img src="https://img.shields.io/github/stars/nikondrat/cctx-mcp?style=flat-square&logo=github&logoColor=white&label=Stars&labelColor=1E293B&color=3B82F6" alt="Stars"></a> <a href="https://github.com/nikondrat/cctx-mcp/actions"><img src="https://img.shields.io/github/actions/workflow/status/nikondrat/cctx-mcp/ci.yml?style=flat-square&logo=githubactions&logoColor=white&label=CI&labelColor=1E293B&color=3B82F6" alt="CI"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-3B82F6?style=flat-square&labelColor=1E293B" alt="License"></a> <img src="https://img.shields.io/badge/Status-Beta-3B82F6?style=flat-square&labelColor=1E293B" alt="Beta"> </p>
Overview
CCTX-MCP — Code ConTeXt via Model Context Protocol. An MCP server that gives AI agents a structured view of source code without reading entire files.
Built for Claude, Cursor, OpenCode, and any MCP-compatible AI coding tool.
<p align="center"> <code>uvx cctx-mcp</code> · <a href="#%EF%B8%8F-tools">Tools</a> · <a href="#-token-savings">Savings</a> · <a href="#-install">Install</a> </p>
Quick Start
uvx cctx-mcp
Add to your MCP client config:
{
"mcpServers": {
"cctx-mcp": {
"command": "uvx",
"args": ["cctx-mcp"]
}
}
}
Tools
Code Analysis
| Tool | Returns | Replaces | Savings |
|---|---|---|---|
smart_read |
symbol hierarchy, deps, docs, line counts | cat + manual parsing |
~87% |
find_symbols |
symbol locations by name or type | grep -r + file reads |
~99% |
get_dependencies |
all imports of a file in one shot | grep ^import |
~96% |
trace_calls |
every call site with file + line | grep across repo |
~90% |
analyze_project |
language breakdown, file counts, tree | find + wc |
~98% |
get_symbol_summaries |
AI semantic descriptions per symbol | reading implementation | ~85% |
Git & Commits
| Tool | Returns | Replaces | Savings |
|---|---|---|---|
compact_change_intelligence |
structured git diff with intent cues | git diff + git status |
~75% |
draft_commit |
AI-generated conventional commit message | writing from scratch | ~90% |
approve_commit_draft |
executes the commit after review | git add + git commit |
— |
Observability
get_config · get_health · get_version
Token Savings
| Operation | Native | With CCTX-MCP | Savings |
|---|---|---|---|
| Read 500-line file | ~1,500 tokens | ~200 tokens | 87% |
| Find function across project | ~5,000 tokens | ~50 tokens | 99% |
| Understand imports | ~800 tokens | ~30 tokens | 96% |
| Analyze project structure | ~10,000 tokens | ~150 tokens | 98% |
| Git change summary | ~3,000 tokens | ~750 tokens | 75% |
Typical session: 80%+ aggregate savings.
Supported Languages
<p> <img src="https://img.shields.io/badge/Swift-FA7343?style=flat-square&logo=swift&logoColor=white" alt="Swift"> <img src="https://img.shields.io/badge/Python-3776AB?style=flat-square&logo=python&logoColor=white" alt="Python"> <img src="https://img.shields.io/badge/TypeScript-3178C6?style=flat-square&logo=typescript&logoColor=white" alt="TypeScript"> <img src="https://img.shields.io/badge/JavaScript-F7DF1E?style=flat-square&logo=javascript&logoColor=black" alt="JavaScript"> <img src="https://img.shields.io/badge/Rust-000000?style=flat-square&logo=rust&logoColor=white" alt="Rust"> <img src="https://img.shields.io/badge/Go-00ADD8?style=flat-square&logo=go&logoColor=white" alt="Go"> <img src="https://img.shields.io/badge/Dart-0175C2?style=flat-square&logo=dart&logoColor=white" alt="Dart"> </p>
Powered by tree-sitter AST — each language has a dedicated parser.
Install
uvx (recommended)
{
"mcpServers": {
"cctx-mcp": {
"command": "uvx",
"args": ["cctx-mcp"]
}
}
}
pip
pip install cctx-mcp
{
"mcpServers": {
"cctx-mcp": {
"command": "python",
"args": ["-m", "code_context.server"]
}
}
}
source
git clone https://github.com/nikondrat/cctx-mcp.git
cd cctx-mcp
uv sync
uv run python -m code_context.server --skip-index
Configuration
| Variable | Default | Description |
|---|---|---|
CC_OLLAMA_URL |
http://localhost:11434 |
Ollama server address |
CC_OPENROUTER_API_KEY |
— | Remote inference key |
CC_LLM_ROUTER |
local-first |
local-first, local-only, remote-first, remote-only |
CC_COMMIT_MODEL |
gemma4:latest |
Local model for commit drafting |
CC_EMBED_MODEL |
nomic-embed-text |
Local model for embeddings |
CC_SEMANTIC_SUMMARIES |
1 |
Enable AI symbol summaries |
Development
uv sync
uv run pytest tests/ -v
Debug Logging
Tool calls are logged to ~/.code-context-cache/debug.jsonl with args, result preview, latency, and status:
tail -f ~/.code-context-cache/debug.jsonl
Set CC_DEBUG_LOG env var to change the log path.
PRs welcome. Open issues.
License
MIT — free for any use.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。