linuxdo-mcp

# linuxdo-mcp #### This post uses community public welfare promotion, which meets the promotion requirements. I declare and follow the following contents required by the community: * **My project is free to use, without charging (implicit charging, sponsorship) part:** Yes * **My post has been marked with #公益推广 tag:** Yes * **My project belongs to personal project, not related to company or commercial institution:** Yes * **My project does not have QQ, TG and other group guides:** Yes * **My project does not have non-operational necessary website guides:** Yes * **My project does not promote for others, AFF:** Yes * **My project has no associated commercial projects:** Yes * **My site has login, and has accessed LINUX DO Connect:** No * **The project introduction in my post, AI generated, polished content has been screenshot and sent out:** Yes * **The above choices I promise are permanently effective, and accept the supervision of the community and old friends:** Yes Search / Read [linux.do](https://linux.do)(Discourse Forum) with MCP Server. Use [curl_cffi](https://github.com/lexiforest/curl_cffi) to simulate Chrome TLS fingerprint to bypass Cloudflare, access trusted level restricted content with login cookie. ## Tool List | Tool | Return | Description | |------|------|------| | `whoami()` | JSON | Current cookie corresponding login user and trust level | | `search(query, page=1, pages=1)` | JSON | Full search, `query` supports Discourse advanced syntax | | `get_topic(topic_id, posts=5)` | JSON | Topic details + first N floor posts | | `list_categories()` | JSON | All categories, including respective topic count `topic_count`, post count `post_count` | | `category_topics(category_id, page=1)` | JSON | Topic list under specified category (about 30 per page), including total topic count of the category | | `list_tags()` | JSON | All tags and respective topic count `count` | | `tag_topics(tag, page=1)` | JSON | Topic list under specified tag | | `user_info(username)` | JSON | User information: trust level, title, post count, like count, registration/online time | | `latest_topics(page=1)` | JSON | Latest topics on the homepage | | `top_topics(period="weekly", page=1)` | JSON | Popular topics, period: daily/weekly/monthly/quarterly/yearly/all | | `user_actions(username, limit=20)` | JSON | User's posting/reply activities (including summary and link) | | `format_search(query, page=1, pages=1)` | Markdown | Same as search, directly returns finished Markdown (title + URL + summary) | | `format_topic(topic_id, posts=20)` | Markdown | Same as get_topic, directly returns finished Markdown (source header + floor table) | - `format_*` tools return composed Markdown strings, which can be directly displayed by clients; others return structured JSON. - Advanced search syntax: `order:latest`, `#category`, `@user`, `tags:tag`, `after:2025-01-01`, `in:title`, etc. ## Prerequisites - Install [uv](https://docs.astral.sh/uv/)(provides `uvx`). - Get your own cookie: log in to linux.do on the browser → F12 → Application → Cookies → `https://linux.do` → copy the value of `_t`. Only `_t` is needed, **not** `cf_clearance`. ## Configuration (copy to your MCP client) No need to download code, `uvx` will automatically pull and run from the repository. Add the following to your MCP client configuration, fill in **your own** `_t`: ```json { "mcpServers": { "linuxdo": { "command": "uvx", "args": ["--from", "git+https://github.com/mrsxs/linuxdo-mcp", "linuxdo-mcp"], "env": { "LINUXDO_COOKIE": "_t=your_token_value" } } } } ``` - **Claude Code**: `claude mcp add-json linuxdo '<above content>'`, or write to `.mcp.json` / settings. - **Cursor / Claude Desktop / Cline**: paste into respective MCP configuration files. > ⚠️ `_t` is your linux.do login credential, only fill in your own local configuration, **do not share with others**. ## Local Running (Development) ```bash export LINUXDO_COOKIE='_t=...' uvx --from . linuxdo-mcp # or uv run src/linuxdo_mcp/server.py ``` ## Remarks - Cookie expiration will return 401/403, re-export `_t` is ok. - Occasionally blocked by Cloudflare will automatically retry 3 times; still failed can set `LINUXDO_IMPERSONATE=chrome131` (or `chrome124`) to change fingerprint. - All requests are read-only GET, no write operations.