Skip to content
Houtini.
Contact
All tools · AI & Developer Tools

Houtini LM

Offload the heavy reading and bulk analysis to a local LLM, on hardware you own. No quota burn.

81 stars 1.4k/mo Updated 13 days ago @houtini/lm
How to install Need help? Talk to us
What it does

In practice.

Wires a local Qwen Coder Next instance (120k context) into the Claude workflow. Use it for log triage, voice extraction, batch summarisation - anything where the work is bounded and you'd rather not pay per token.

Local LLM 120k context Bounded tasks Cost control
How to install

Three steps. Five minutes.

Houtini LM runs as an MCP server. You install it once, point Claude at it via a small config file, and it's available in every conversation from that point on. The full primer is at how to add an MCP server to Claude Desktop; what follows is the short version.

  1. 1

    Install the package

    Open a terminal and run the command below. It installs Houtini LM globally so Claude can find it from any directory.

    npm i -g @houtini/lm
  2. 2

    Add it to your Claude config

    In Claude Desktop, open Settings → Developer → Edit Config. Add an entry under mcpServers pointing at @houtini/lm. The full primer (with the JSON snippet you can copy) is at how to add an MCP server to Claude Desktop.

  3. 3

    Restart Claude

    Close Claude Desktop completely and reopen it. The new MCP shows up in the tool list at the bottom of the message box. If it doesn't, the linked primer above has the troubleshooting checklist.

More

Source, package, docs.

GitHub github.com/houtini-ai/lm 81 stars · MIT npm @houtini/lm 1.4k/mo
If you want a hand

Want it doing something specific to you?

Houtini LM is free, MIT-licensed, runs as it is. If you want it pointed at your own data, wired to your CRM, or built around a workflow this one doesn't cover, that's a conversation. We build bespoke MCPs for clients. Forty-five minutes, no pitch deck.

Talk to us How custom MCP builds work