Redd-Archiver MCP

Install

Works in

• Claude Desktop
• Claude Code
• SourceCraft
• Koda
• Coddy

Manual install

{
  "mcpServers": {
    "reddarchiver": {
      "command": "uv",
      "args": ["--directory", "/path/to/mcp_server", "run", "python", "server.py"],
      "env": { "REDDARCHIVER_API_URL": "http://localhost:5000" }
    }
  }
}

What it does

Redd-Archiver turns compressed Reddit, Voat and Ruqqus dumps into a browsable archive. The tool loads data into PostgreSQL and builds HTML pages from the database with a mobile-first layout, full-text search and eleven color themes. It detects the platform from the file extension: for Reddit that is `.zst` JSON Lines from Pushshift or Arctic Shift, for Voat SQL dumps, for Ruqqus `.7z`. The archive is served in three modes. Static exports once and can be hosted anywhere, including GitHub Pages or a USB stick. Hybrid, the default, adds a search server on nginx, Flask and PostgreSQL. Dynamic renders pages from the database with no export step. Monthly Arctic Shift dumps keep it current: `reddarc.py --update` applies them to an existing archive and `--update-all` walks a folder of dumps. For agents the repository includes a FastMCP server. It builds 29 tools from the OpenAPI description of the archive's REST interface, exposes five resources (stats, top posts, communities, search help) and caps response size so the context does not overflow. The agent searches posts and comments, looks up users and computes aggregates over your archive.

Who it is for. For researchers, archivists and analysts who keep their own copy of Reddit communities and want to question it through Claude.

Example request

Find posts in the privacy subreddit archive from 2022 about password managers, show the five most discussed and summarize the comments

MCP

Transport

stdio

Authentication

not required

Limitations

Python 3.10 or newer and PostgreSQL are required. The tool does not download data for you: Reddit dumps come from Arctic Shift and Academic Torrents, Voat and Ruqqus from the Internet Archive. Dumps are large; see the performance documentation for memory and disk requirements.

The MCP server works only on top of a running archive search server: without it the tools get no data.

The project is released under the Unlicense. Rights to the community material belong to its authors and platforms, so weigh the legal terms before publishing an archive.

How to disable. Remove the reddarchiver block from your client settings and restart it. Stop the services with `docker compose stop mcp-server`, or the whole stack with `docker compose down`. The PostgreSQL database holding the archive is stored separately; delete it if you no longer need the archive.

README in short

The description opens with a platform table with data volumes and an explanation of the three serving modes. Monthly dump updates, a quick start with Tor or HTTPS deployment, an MCP server section and a list of features for readers and operators follow.