Configuration

MUR's configuration lives at ~/.mur/config.yaml. Run mur init for an interactive setup wizard.

Full Configuration Reference

# Semantic search / embedding
embedding:
  provider: ollama              # ollama | openai | gemini | anthropic
  model: qwen3-embedding:0.6b  # see Embedding Models below
  dimensions: 1024              # must match model (set automatically by mur init)
  ollama_endpoint: http://localhost:11434

# Pattern extraction LLM
llm:
  provider: anthropic           # anthropic | openai | gemini | ollama
  model: claude-sonnet-4-20250514
  api_key_env: ANTHROPIC_API_KEY
  # openai_url: https://openrouter.ai/api/v1  # for OpenRouter

Model Registry (agents & cost router)

The models that agents and the cost router use live in a separate registry at ~/.mur/models.yaml, not in config.yaml. Manage it with the mur model commands rather than editing the file by hand:

mur model add gpt5 --provider openai --model gpt-5.2 --secret env:OPENAI_API_KEY
mur model list

Pricing (input/output USD-per-1k) and context window are auto-filled from the models.dev catalog. API keys are stored as SecretRefs (env: / keychain: / file: / cmd:), never in plaintext. The MUR Hub desktop app's Model Library offers a GUI: connect a cloud provider (key → macOS Keychain) or auto-detect a local runtime (Ollama / MLX / LM Studio), then pick which discovered models to register.

Embedding Models

Ollama (Free, Local — Recommended)

No API key needed, runs entirely on your machine:

# Pull your chosen model first
ollama pull qwen3-embedding:0.6b

embedding:
  provider: ollama
  model: qwen3-embedding:0.6b
  dimensions: 1024

OpenAI

Low cost (~$0.001 per 200 patterns):

embedding:
  provider: openai
  model: text-embedding-3-small
  dimensions: 1536
  api_key_env: OPENAI_API_KEY

After changing embedding config, rebuild the index:

mur reindex

LLM Providers

MUR uses an LLM for pattern extraction (mur learn extract) and contradiction detection (mur feedback auto).

Cloud providers

llm:
  provider: anthropic           # or openai, gemini
  model: claude-sonnet-4-20250514
  api_key_env: ANTHROPIC_API_KEY

OpenRouter (access to many models)

llm:
  provider: openai
  model: google/gemini-2.5-flash
  api_key_env: OPENROUTER_API_KEY
  openai_url: https://openrouter.ai/api/v1

Ollama (free, local)

llm:
  provider: ollama
  model: llama3.2:3b

File Locations

See Also

• Semantic Search — Deep dive into search configuration
• Commands — Full command reference