# gcontext-mcp

The local MCP connector for [gcontext](https://gcontext.ai). Runs on your machine as a stdio MCP server, bridging your AI agent to your gcontext workspace in the cloud.

- **Structure** (files, folders, types, secret names) lives in the cloud
- **Secret values** and **script execution** stay local, on your machine

## Install

Add to Claude Code:

```bash
claude mcp add-json gcontext-personal '{"type":"stdio","command":"uvx","args":["gcontext-mcp"],"env":{"GCONTEXT_TOKEN":"<your token>"}}'
```

Or run directly:

```bash
uvx gcontext-mcp
```

## Environment variables

| Variable | Required | Description |
|---|---|---|
| `GCONTEXT_TOKEN` | yes | Your workspace token (from app.gcontext.ai) |
| `GCONTEXT_API_URL` | no | API endpoint (default: `https://app.gcontext.ai`) |

## Self-hosting

If you self-host the gcontext backend, point `GCONTEXT_API_URL` to your instance:

```json
{"type":"stdio","command":"uvx","args":["gcontext-mcp"],"env":{"GCONTEXT_TOKEN":"<token>","GCONTEXT_API_URL":"http://localhost:5179"}}
```

See [self-hosting docs](https://gcontext.ai/docs) for backend setup.

## What this connector does

The connector exposes MCP tools for reading, writing, searching, and managing files in your gcontext workspace. Secret values are stored locally (in `~/.gcontext/.env`) and injected at script runtime without ever leaving your machine.

## License

MIT