@houtini/lm Houtini LM - Offload Tasks from Claude Code to Your Local LLM Server (LM Studio / Ollama) or a Cloud API

I built this because I kept leaving Claude Code running overnight on big refactors and the token bill was painful. A huge chunk of that spend goes on bounded tasks any decent model handles fine - generating boilerplate, code review, commit messages, format conversion. Stuff that doesn't need Claude's reasoning or tool access.

Houtini LM connects Claude Code to a local LLM on your network - or any OpenAI-compatible API. Claude keeps doing the hard work - architecture, planning, multi-file changes - and offloads the grunt work to whatever cheaper model you've got running. Free. No rate limits. Private.

I wrote a full walkthrough of why I built this and how I use it day to day.

How it works

Claude Code (orchestrator)
   |
   |-- Complex reasoning, planning, architecture --> Claude API (your tokens)
   |
   +-- Bounded grunt work --> houtini-lm --HTTP/SSE--> Your local LLM (free)
       . Boilerplate & test stubs          Qwen, Llama, Nemotron, GLM...
       . Code review & explanations        LM Studio, Ollama, vLLM, llama.cpp
       . Commit messages & docs            DeepSeek, Groq, Cerebras (cloud)
       . Format conversion
       . Mock data & type definitions
       . Embeddings for RAG pipelines

Claude's the architect. Your local model's the drafter. Claude QAs everything.

Every response comes back with performance stats - TTFT, tokens per second, generation time - so you can actually see what your local hardware is doing. The session footer tracks cumulative offloaded tokens across every call.

Quick start

Claude Code

claude mcp add houtini-lm -- npx -y @houtini/lm

That's it. If LM Studio's running on localhost:1234 (the default), Claude can start delegating straight away.

LLM on a different machine

I've got a GPU box on my local network running Qwen 3 Coder Next in LM Studio. If you've got a similar setup, point the URL at it:

claude mcp add houtini-lm -e LM_STUDIO_URL=http://192.168.1.50:1234 -- npx -y @houtini/lm

Cloud APIs

Works with anything speaking the OpenAI format. DeepSeek at twenty-eight cents per million tokens, Groq for speed, Cerebras if you want three thousand tokens per second - whatever you fancy:

claude mcp add houtini-lm \
  -e LM_STUDIO_URL=https://api.deepseek.com \
  -e LM_STUDIO_PASSWORD=your-key-here \
  -- npx -y @houtini/lm

Claude Desktop

Drop this into your claude_desktop_config.json:

{
  "mcpServers": {
    "houtini-lm": {
      "command": "npx",
      "args": ["-y", "@houtini/lm"],
      "env": {
        "LM_STUDIO_URL": "http://localhost:1234"
      }
    }
  }
}

Model discovery

This is where things get interesting. At startup, houtini-lm queries your LLM server for every model available - loaded and downloaded - then looks each one up on HuggingFace's free API to pull metadata: architecture, licence, download count, pipeline type. All of that gets cached in a local SQLite database (~/.houtini-lm/model-cache.db) so subsequent startups are instant.

The result is that houtini-lm actually knows what your models are good at. Not just the name - the capabilities, the strengths, what tasks to send where. If you've got Nemotron loaded but a Qwen Coder sitting idle, it'll flag that. If someone on a completely different setup loads a Mistral model houtini-lm has never seen before, the HuggingFace lookup auto-generates a profile for it.

Run list_models and you get the full picture:

Loaded models (ready to use):

  nvidia/nemotron-3-nano
    type: llm, arch: nemotron_h_moe, quant: Q4_K_M, format: gguf
    context: 200,082 (max 1,048,576), by: nvidia
    Capabilities: tool_use
    NVIDIA Nemotron: compact reasoning model optimised for step-by-step logic
    Best for: analysis tasks, code bug-finding, math/science questions
    HuggingFace: text-generation, 1.7M downloads, MIT licence

Available models (downloaded, not loaded):

  qwen3-coder-30b-a3b-instruct
    type: llm, arch: qwen3moe, quant: BF16, context: 262,144
    Qwen3 Coder: code-specialised model with agentic capabilities
    Best for: code generation, code review, test stubs, refactoring
    HuggingFace: text-generation, 12.9K downloads, Apache-2.0

For models we know well - Qwen, Nemotron, Granite, LLaMA, GLM, GPT-OSS - there's a curated profile built in with specific strengths and weaknesses. For everything else, the HuggingFace lookup fills the gaps. Cache refreshes every 7 days. Zero friction - sql.js is pure WASM, no native dependencies, no build tools needed.

What gets offloaded

Delegate to the local model - bounded, well-defined tasks:

Task	Why it works locally
Generate test stubs	Clear input (source), clear output (tests)
Explain a function	Summarisation doesn't need tool access
Draft commit messages	Diff in, message out
Code review	Paste full source, ask for bugs
Convert formats	JSON to YAML, snake_case to camelCase
Generate mock data	Schema in, data out
Write type definitions	Source in, types out
Structured JSON output	Grammar-constrained, guaranteed valid
Text embeddings	Semantic search, RAG pipelines
Brainstorm approaches	Doesn't commit to anything

Keep on Claude - anything that needs reasoning, tool access, or multi-step orchestration:

• Architectural decisions
• Reading/writing files
• Running tests and interpreting results
• Multi-file refactoring plans
• Anything that needs to call other tools

The tool descriptions are written to nudge Claude into planning delegation at the start of large tasks, not just using it when it happens to think of it.

Performance tracking

Every response includes a footer with real performance data - computed from the SSE stream, not from any proprietary API:

Model: zai-org/glm-4.7-flash | 125->430 tokens | TTFT: 678ms, 48.7 tok/s, 12.5s
Session: 8,450 tokens offloaded across 14 calls

The discover tool shows per-model averages across the session:

Performance (this session):
  nvidia/nemotron-3-nano: 6 calls, avg TTFT 234ms, avg 45.2 tok/s
  zai-org/glm-4.7-flash: 8 calls, avg TTFT 678ms, avg 48.7 tok/s

In practice, Claude delegates more aggressively the longer a session runs. After about 5,000 offloaded tokens, it starts hunting for more work to push over. Reinforcing loop.

Model routing

If you've got multiple models loaded (or downloaded), houtini-lm picks the best one for each task automatically. Each model family has per-family prompt hints - temperature, output constraints, and think-block flags - so GLM gets told "no preamble, no step-by-step reasoning" while Qwen Coder gets a low temperature for focused code output.

The routing scores loaded models against the task type (code, chat, analysis, embedding). If the best loaded model isn't ideal for the task, you'll see a suggestion in the response footer pointing to a better downloaded model. No runtime model swapping - model loading takes minutes, so houtini-lm suggests rather than blocks.

Supported model families with curated prompt hints: GLM-4, Qwen3 Coder, Qwen3, LLaMA 3, Nemotron, Granite, GPT-OSS, Nomic Embed. Unknown models get sensible defaults.

Tools

chat

The workhorse. Send a task, get an answer. The description includes planning triggers that nudge Claude to identify offloadable work when it's starting a big task.

Parameter	Required	Default	What it does
message	yes	-	The task. Be specific about output format.
system	no	-	Persona - "Senior TypeScript dev" not "helpful assistant"
temperature	no	0.3	0.1 for code, 0.3 for analysis, 0.7 for creative
max_tokens	no	2048	Lower for quick answers, higher for generation
json_schema	no	-	Force structured JSON output conforming to a schema

custom_prompt

Three-part prompt: system, context, instruction. Keeping them separate prevents context bleed - consistently outperforms stuffing everything into one message, especially with local models. I tested this properly one weekend - took the same batch of review tasks and ran them both ways. Splitting things into three parts won every round.

Parameter	Required	Default	What it does
instruction	yes	-	What to produce. Under 50 words works best.
system	no	-	Persona + constraints, under 30 words
context	no	-	Complete data to analyse. Never truncate.
temperature	no	0.3	0.1 for review, 0.3 for analysis
max_tokens	no	2048	Match to expected output length
json_schema	no	-	Force structured JSON output

code_task

Built for code analysis. Pre-configured system prompt with temperature and output constraints tuned per model family via the routing layer.

Parameter	Required	Default	What it does
code	yes	-	Complete source code. Never truncate.
task	yes	-	"Find bugs", "Explain this", "Write tests"
language	no	-	"typescript", "python", "rust", etc.
max_tokens	no	2048	Match to expected output length

embed

Generate text embeddings via the OpenAI-compatible /v1/embeddings endpoint. Requires an embedding model to be available - Nomic Embed is a solid choice. Returns the vector, dimension count, and usage stats.

Parameter	Required	Default	What it does
input	yes	-	Text to embed
model	no	auto	Embedding model ID

discover

Health check. Returns model name, context window, latency, capability profile, and cumulative session stats including per-model performance averages. Call before delegating if you're not sure the LLM's available.

list_models

Lists everything on the LLM server - loaded and downloaded - with full metadata: architecture, quantisation, context window, capabilities, and HuggingFace enrichment data. Shows capability profiles describing what each model is best at, so Claude can make informed delegation decisions.

Structured JSON output

Both chat and custom_prompt accept a json_schema parameter that forces the response to conform to a JSON Schema. LM Studio uses grammar-based sampling to guarantee valid output - no hoping the model remembers to close its brackets.

{
  "json_schema": {
    "name": "code_review",
    "schema": {
      "type": "object",
      "properties": {
        "issues": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "line": { "type": "number" },
              "severity": { "type": "string" },
              "description": { "type": "string" }
            },
            "required": ["line", "severity", "description"]
          }
        }
      },
      "required": ["issues"]
    }
  }
}

Getting good results from local models

Qwen, Llama, Nemotron, GLM - they score brilliantly on coding benchmarks now. The gap between a good and bad result is almost always prompt quality, not model capability. I've spent a fair bit of time on this.

Send complete code. Local models hallucinate details when you give them truncated input. If a file's too large, send the relevant function - not a snippet with ... in the middle.

Be explicit about output format. "Return a JSON array" or "respond in bullet points" - don't leave it open-ended. Smaller models need this.

Set a specific persona. "Expert Rust developer who cares about memory safety" gets noticeably better results than "helpful assistant."

State constraints. "No preamble", "reference line numbers", "max 5 bullet points" - tell the model what not to do as well as what to do.

Include surrounding context. For code generation, send imports, types, and function signatures - not just the function body.

One call at a time. If your LLM server runs a single model, parallel calls queue up and stack timeouts. Send them sequentially.

Think-block stripping

Some models - GLM Flash, Nemotron, and others - always emit <think>...</think> reasoning blocks before the actual answer. Houtini-lm strips these automatically so Claude gets clean output without wasting time parsing the model's internal chain-of-thought. You still get the benefit of the reasoning (better answers), just without the noise.

Configuration

Variable	Default	What it does
LM_STUDIO_URL	http://localhost:1234	Base URL of the OpenAI-compatible API
LM_STUDIO_MODEL	(auto-detect)	Model identifier - leave blank to use whatever's loaded
LM_STUDIO_PASSWORD	(none)	Bearer token for authenticated endpoints
LM_CONTEXT_WINDOW	100000	Fallback context window if the API doesn't report it

Compatible endpoints

Works with anything that speaks the OpenAI /v1/chat/completions API:

What	URL	Notes
LM Studio	http://localhost:1234	Default, zero config. Rich metadata via v0 API.
Ollama	http://localhost:11434	Set LM_STUDIO_URL
vLLM	http://localhost:8000	Native OpenAI API
llama.cpp	http://localhost:8080	Server mode
DeepSeek	https://api.deepseek.com	28c/M input tokens
Groq	https://api.groq.com/openai	~750 tok/s
Cerebras	https://api.cerebras.ai	~3000 tok/s
Any OpenAI-compatible API	Any URL	Set URL + password

Streaming and timeouts

All inference uses Server-Sent Events streaming. Tokens arrive incrementally, keeping the connection alive. If generation takes longer than 55 seconds, you get a partial result instead of a timeout error - the footer shows TRUNCATED when this happens.

The 55-second soft timeout exists because the MCP SDK has a hard ~60s client-side timeout. Without streaming, any response that took longer than 60 seconds just vanished. Not ideal.

Architecture

index.ts          Main MCP server - tools, streaming, session tracking
model-cache.ts    SQLite-backed model profile cache (sql.js / WASM)
                  Auto-profiles models via HuggingFace API at startup
                  Persists to ~/.houtini-lm/model-cache.db

Inference:        POST /v1/chat/completions  (OpenAI-compatible, works everywhere)
Model metadata:   GET  /api/v0/models        (LM Studio, falls back to /v1/models)
Embeddings:       POST /v1/embeddings        (OpenAI-compatible)

Development

git clone https://github.com/houtini-ai/lm.git
cd lm
npm install
npm run build

Licence

MIT