Configuration /config

Config file

All preferences are stored in ~/.config/muxd/config.json. You can edit this file directly or use /config set from the TUI.

JSON keys vs /config set keys: config.json uses snake_case (underscores), while /config set uses dot notation. For example, model_compact in config.json is model.compact in /config set. The two formats are interchangeable — use whichever you prefer.

Here's a comprehensive real-world example showing all the main fields:

{
  "footer_tokens": true,
  "footer_cost": true,
  "footer_cwd": true,
  "footer_session": true,
  "footer_keybindings": true,

  "model": "zai/glm-5.1",
  "model_compact": "zai/glm-5.1",
  "model_consult": "accounts/fireworks/models/kimi-k2p5",
  "model_title": "accounts/fireworks/models/gpt-oss-120b",
  "model_tags": "accounts/fireworks/models/gpt-oss-120b",
  "provider": "zai",

  "zai_api_key": "your-zai-key",
  "zai_coding_plan": true,
  "fireworks_api_key": "your-fireworks-key",
  "brave_api_key": "your-brave-key",
  "textbelt_api_key": "your-textbelt-key",

  "scheduler_allowed_tools": "sms_send",

  "daemon_auth_token": "your-daemon-token",

  "hub_url": "http://192.168.1.50:4097",
  "hub_node_token": "your-hub-token",
  "hub_node_name": "my-desktop",

  "tier1_percent": 0.65,
  "tier3_percent": 0.85,
  "tail_token_percent": 0.3,

  "mobile_models": [
    { "id": "zai/glm-5.1", "display": "GLM 5.1", "provider": "Z.AI" },
    { "id": "accounts/fireworks/models/kimi-k2p5", "display": "Kimi k2p5", "provider": "Kimi" }
  ]
}

Only set the keys you need — omitted keys use defaults. The file is created automatically on first run.

/config Subcommands

/config set Keys

Daemon

Use 0.0.0.0 to make the daemon accessible from other devices on your LAN (e.g. the muxd mobile app):

/config set daemon.bind_address 0.0.0.0

Or via environment variable:

# Linux / macOS
export MUXD_BIND=0.0.0.0
muxd-daemon

# Windows (PowerShell)
$env:MUXD_BIND = "0.0.0.0"
muxd-daemon

Footer Display

Boolean values accept: true/false, on/off, yes/no, 1/0 (case-insensitive).

Preset emoji names: devil, imp, fire, rocket, skull, robot, ghost, alien, bolt, star, diamond, snake, cat, dog, heart. Use none to remove.

/config set footer.emoji devil
/config set footer.emoji none

Or use the interactive picker: /emoji

Theme

Set directly:

/config set theme.spinner mini-dot

Or use the interactive picker:

/spinner

Hub

Settings for running the hub coordinator (muxd-hub).

Mobile

The iOS app has a model picker that fetches its list from the daemon via GET /api/mobile/config. By default it returns a built-in list of Claude and OpenAI models. Override it by adding a mobile_models array directly to ~/.config/muxd/config.json — this key cannot be set via /config set.

{
  "mobile_models": [
    { "id": "zai/glm-5.1", "display": "GLM 5.1", "provider": "Z.AI" },
    { "id": "accounts/fireworks/models/kimi-k2p5", "display": "Kimi k2p5", "provider": "Kimi" },
    { "id": "claude-sonnet-4-6", "display": "Claude Sonnet 4.6", "provider": "Claude" },
    { "id": "gpt-4o", "display": "GPT-4o", "provider": "OpenAI" }
  ]
}

Each entry requires three fields:

View the current setting with /config mobile. When mobile_models is not set, the app uses the built-in default list.

Model

All five accept full model IDs with or without provider prefix:

/config set model zai/glm-5.1
/config set model.compact zai/glm-5.1
/config set model.consult accounts/fireworks/models/kimi-k2p5
/config set model.title accounts/fireworks/models/gpt-oss-120b
/config set model.tags accounts/fireworks/models/gpt-oss-120b

Or set directly in config.json using the JSON field names.

Node

Settings for registering a daemon as a node with an existing hub.

See the Hub setup guide for the full connection flow.

Providers

Set API keys via /config set or environment variables. Resolution order: environment variable > /config set.

Or use environment variables directly:

export ANTHROPIC_API_KEY="sk-ant-..."
export OPENAI_API_KEY="sk-..."
export MISTRAL_API_KEY="..."
export GROK_API_KEY="xai-..."
export ZAI_API_KEY="..."
export GOOGLE_API_KEY="..."
export FIREWORKS_API_KEY="..."
export DEEPINFRA_API_KEY="..."
export OLLAMA_URL="http://localhost:11434"

Scheduler

Note: Scheduled agent tasks (created via schedule_task or /schedule add-task) bypass the scheduler.allowed_tools allowlist. The spawned agent enforces its own tool policy at runtime.

SMS

/config set textbelt.api_key your-api-key

Get an API key at textbelt.com. Enables the sms_send, sms_status, and sms_schedule tools.

Tools

Boolean values accept: true/false, on/off, yes/no, 1/0 (case-insensitive).

Permission modes:

• ask: prompt before bash and file edits (default)
• accept-edits: auto-approve in-project file edits, still ask for bash
• full-auto: auto-approve most actions, but still blocks sensitive paths

/permissions ask
/permissions accept-edits
/permissions full-auto

Model Aliases

Use short aliases instead of full model IDs:

The anthropic/ prefix is optional: anthropic/claude-sonnet and claude-sonnet are equivalent.

/config set model claude-opus

Provider Prefixes

Use provider/model syntax to specify a model from any provider:

The fireworks/ prefix is shorthand — the provider automatically prepends accounts/fireworks/models/. Both fireworks/kimi-k2p5 and accounts/fireworks/models/kimi-k2p5 work.

Auto-Detection

When no prefix is given, muxd detects the provider from the model name:

Provider Model Examples

/config set model claude-sonnet                                              # Anthropic (alias)
/config set model gpt-4o                                                     # OpenAI (auto-detected)
/config set model mistral-large-latest                                       # Mistral (auto-detected)
/config set model grok-3                                                     # Grok / xAI (auto-detected)
/config set model zai/glm-5.1                                                # Z.AI (prefix)
/config set model fireworks/kimi-k2p5                                        # Fireworks (prefix)
/config set model accounts/fireworks/models/kimi-k2p5                       # Fireworks (full path)
/config set model deepinfra/meta-llama/Meta-Llama-3.1-70B-Instruct          # DeepInfra (prefix)
/config set model google/gemini-2.0-flash                                    # Google (prefix)
/config set model llama3.3:latest                                            # Ollama (auto-detected by ":")

Pricing

muxd tracks estimated costs per session.

Built-in pricing

Custom pricing

Override or add pricing in ~/.config/muxd/pricing.json:

{
  "my-custom-model": {
    "input": 1.0,
    "output": 5.0
  }
}

Values are dollars per million tokens. Custom entries merge with defaults.

Data Directory