Configuration file reference

The Operator configuration file lives at ~/.operator/config.json (or $OPERATOR_HOME/config.json if the environment variable is set). It is loaded once at startup; most fields can also be overridden at runtime via environment variables.

Full example config.json

config.json

{
  "agents": {
    "defaults": {
      "workspace": "~/.operator/workspace",
      "restrict_to_workspace": true,
      "model_name": "gpt4",
      "max_tokens": 8192,
      "temperature": 0.7,
      "max_tool_iterations": 20
    }
  },
  "model_list": [
    {
      "model_name": "gpt4",
      "model": "openai/gpt-5.2",
      "api_key": "sk-your-openai-key",
      "api_base": "https://api.openai.com/v1"
    },
    {
      "model_name": "claude-sonnet-4.6",
      "model": "anthropic/claude-sonnet-4.6",
      "api_key": "sk-ant-your-key"
    },
    {
      "model_name": "gemini",
      "model": "antigravity/gemini-2.0-flash",
      "auth_method": "oauth"
    },
    {
      "model_name": "loadbalanced-gpt4",
      "model": "openai/gpt-5.2",
      "api_key": "sk-key1",
      "api_base": "https://api1.example.com/v1"
    },
    {
      "model_name": "loadbalanced-gpt4",
      "model": "openai/gpt-5.2",
      "api_key": "sk-key2",
      "api_base": "https://api2.example.com/v1"
    }
  ],
  "channels": {
    "telegram": {
      "enabled": false,
      "token": "YOUR_TELEGRAM_BOT_TOKEN",
      "allow_from": ["YOUR_USER_ID"],
      "reasoning_channel_id": ""
    },
    "discord": {
      "enabled": false,
      "token": "YOUR_DISCORD_BOT_TOKEN",
      "allow_from": [],
      "group_trigger": { "mention_only": false },
      "reasoning_channel_id": ""
    }
  },
  "tools": {
    "web": {
      "brave": { "enabled": false, "api_key": "YOUR_BRAVE_API_KEY", "max_results": 5 },
      "duckduckgo": { "enabled": true, "max_results": 5 },
      "perplexity": { "enabled": false, "api_key": "pplx-xxx", "max_results": 5 },
      "proxy": ""
    },
    "exec": {
      "enable_deny_patterns": true,
      "custom_deny_patterns": []
    },
    "cron": { "exec_timeout_minutes": 5 },
    "mcp": {
      "enabled": false,
      "servers": {
        "filesystem": {
          "enabled": false,
          "command": "npx",
          "args": ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
        }
      }
    },
    "skills": {
      "registries": {
        "clawhub": {
          "enabled": true,
          "base_url": "https://clawhub.ai"
        }
      }
    }
  },
  "heartbeat": { "enabled": true, "interval": 30 },
  "devices": { "enabled": false, "monitor_usb": true },
  "gateway": { "host": "127.0.0.1", "port": 18790 }
}

`agents`

Global agent settings.

agents.defaults

object

Default parameters applied to every agent unless overridden at the agent level.

Show properties

workspace

string

default:"~/.operator/workspace"

Absolute or ~-prefixed path to the agent’s working directory. The agent reads and writes files here by default. Set OPERATOR_HOME to relocate the base directory; the workspace path is then $OPERATOR_HOME/workspace.

restrict_to_workspace

boolean

default:"true"

When true, the agent cannot read or write files outside the workspace directory. Set to false to allow full filesystem access (use with care).

allow_read_outside_workspace

boolean

default:"false"

When true, the agent can read (but not write) files outside the workspace even if restrict_to_workspace is true. Useful for referencing read-only reference material.

model_name

string

The model_name alias to use by default. Must match a model_name entry in model_list. Example: "gpt4".

Prefer model_name over the deprecated model field. Both are supported during the transition period.

max_tokens

number

default:"32768"

Maximum number of tokens the model may generate in a single response.

temperature

number

default:"null"

Sampling temperature (0.0–2.0). null means the provider default is used. Lower values produce more deterministic output; higher values increase creativity.

max_tool_iterations

number

default:"50"

Maximum number of tool-call/response cycles per agent turn. Prevents runaway loops. Set to 0 for no limit (not recommended).

image_model

string

Optional model alias to use for image-generation tasks. Falls back to model_name if not set.

max_media_size

number

default:"20971520"

Maximum size in bytes for media attachments (images, files) passed to the model. Default is 20 MB (20 × 1024 × 1024).

agents.list

object[]

Optional list of named agents with per-agent overrides. Each entry can override the model, workspace, and skills for a specific agent.

Show properties

string

required

Unique identifier for the agent. Referenced in bindings and --agent CLI flags.

name

string

Human-readable display name.

default

boolean

default:"false"

If true, this agent handles requests that don’t match any binding.

workspace

string

Per-agent workspace directory. Overrides agents.defaults.workspace.

model

string | object

Per-agent model. Either a plain string alias ("gpt4") or an object with primary and fallbacks fields.

Show object form

primary

string

Primary model alias.

fallbacks

string[]

Ordered list of fallback model aliases tried when the primary fails.

skills

string[]

List of skill identifiers pre-loaded for this agent.

subagents

object

Configuration for sub-agent delegation.

Show properties

allow_agents

string[]

Agent IDs this agent may spawn as sub-agents.

model

string | object

Model used for sub-agent tasks.

`model_list`

model_list is the recommended way to configure AI providers. The older providers top-level key is deprecated and will be removed in v2.0. If only providers is present, it is automatically migrated at startup.

An ordered array of model entries. Each entry defines an alias, a provider/model identifier, and credentials. Multiple entries with the same model_name enable automatic round-robin load balancing across them.

{
  "model_list": [
    {
      "model_name": "gpt4",
      "model": "openai/gpt-5.2",
      "api_key": "sk-your-key",
      "api_base": "https://api.openai.com/v1"
    }
  ]
}

model_list[].model_name

string

required

User-facing alias for this model entry. This is the value you set in agents.defaults.model_name. Must be non-empty. Duplicate values across entries are allowed and enable load balancing.

model_list[].model

string

required

Protocol-prefixed model identifier in the form protocol/model-id. The protocol determines which provider backend is used. If no prefix is given, openai is assumed.See the AI provider reference for the full list of supported protocols and example model IDs.

model_list[].api_key

string

API key for authenticating with the provider. Required for all HTTP-based providers unless api_base points to a local unauthenticated server.

model_list[].api_base

string

Override the provider’s default API endpoint URL. Use this to point at a self-hosted deployment, proxy, or alternative region. Each provider has a compiled-in default that applies when this field is omitted.

model_list[].auth_method

string

Authentication method. Set to "oauth" for providers that use OAuth-based login (e.g., antigravity, github-copilot). Omit for API-key–based providers.

model_list[].proxy

string

HTTP/HTTPS/SOCKS5 proxy URL to use for requests to this provider. Example: "http://proxy.corp.example.com:8080".

model_list[].connect_mode

string

Connection mode for CLI-based providers (claude-cli, codex-cli, github-copilot). Valid values: "stdio", "grpc". Defaults to "grpc" for GitHub Copilot.

model_list[].workspace

string

Workspace directory for CLI-based providers (claude-cli, codex-cli). Defaults to "." if not set.

model_list[].rpm

number

Requests-per-minute cap for this entry. When reached, Operator waits before sending the next request.

model_list[].max_tokens_field

string

Override the JSON field name used to pass max_tokens in the API request body. Some providers (e.g., newer OpenAI endpoints) require "max_completion_tokens" instead of "max_tokens".

model_list[].request_timeout

number

Per-request HTTP timeout in seconds. Values <= 0 fall back to the compiled default of 120 s.

Load balancing

Duplicate a model_name across multiple entries to distribute load via round-robin:

{
  "model_list": [
    { "model_name": "gpt4", "model": "openai/gpt-5.2", "api_key": "sk-key1", "api_base": "https://api1.example.com/v1" },
    { "model_name": "gpt4", "model": "openai/gpt-5.2", "api_key": "sk-key2", "api_base": "https://api2.example.com/v1" },
    { "model_name": "gpt4", "model": "openai/gpt-5.2", "api_key": "sk-key3", "api_base": "https://api3.example.com/v1" }
  ]
}

Every call to model gpt4 cycles through the three endpoints in sequence.

`channels`

Messaging platform integrations. All channels are disabled by default. Enable only the channels you need. Every channel config includes the following common fields:

Field	Type	Default	Description
`enabled`	bool	`false`	Enable this channel.
`allow_from`	string[]	`[]`	Allowlist of user/account IDs. Empty means all authenticated users are allowed. Accepts both string and numeric IDs.
`reasoning_channel_id`	string	`""`	If set, internal reasoning/thinking output is forwarded to this channel or thread ID instead of the main conversation.

`channels.telegram`

channels.telegram.enabled

boolean

default:"false"

Enable the Telegram bot.

channels.telegram.token

string

Telegram Bot API token. Obtain from @BotFather.

channels.telegram.base_url

string

Custom Telegram Bot API base URL. Useful when running a local Bot API server.

channels.telegram.proxy

string

HTTP/SOCKS5 proxy for Telegram API requests.

channels.telegram.allow_from

string[]

default:"[]"

Allowlist of Telegram user IDs (numeric or string).

channels.telegram.group_trigger

object

Controls when the bot responds in group chats.

Show properties

mention_only

boolean

default:"false"

Only respond when the bot is @mentioned.

prefixes

string[]

Respond only when a message starts with one of these prefix strings.

channels.telegram.typing

object

Show properties

enabled

boolean

default:"true"

Show “typing…” indicator while the agent is processing.

channels.telegram.placeholder

object

Show properties

enabled

boolean

default:"true"

Send a placeholder message immediately, then edit it with the final response.

text

string

default:"Thinking... 💭"

Text of the placeholder message.

channels.telegram.reasoning_channel_id

string

Forward reasoning output to this chat ID.

`channels.discord`

channels.discord.enabled

boolean

default:"false"

Enable the Discord bot.

channels.discord.token

string

Discord bot token from the Discord Developer Portal.

channels.discord.proxy

string

HTTP/SOCKS5 proxy for Discord API requests.

channels.discord.allow_from

string[]

default:"[]"

Allowlist of Discord user IDs.

channels.discord.group_trigger

object

Show properties

mention_only

boolean

default:"false"

Only respond when the bot is @mentioned in a server channel.

prefixes

string[]

Respond only when a message starts with one of these prefix strings.

channels.discord.reasoning_channel_id

string

Channel ID for reasoning output.

`channels.slack`

channels.slack.enabled

boolean

default:"false"

Enable the Slack bot.

channels.slack.bot_token

string

Slack bot OAuth token (xoxb-…).

channels.slack.app_token

string

Slack app-level token for Socket Mode (xapp-…).

channels.slack.allow_from

string[]

default:"[]"

Allowlist of Slack user IDs.

channels.slack.group_trigger

object

Show properties

mention_only

boolean

default:"false"

Only respond when the bot is mentioned.

channels.slack.reasoning_channel_id

string

Channel ID for reasoning output.

`channels.whatsapp`

channels.whatsapp.enabled

boolean

default:"false"

Enable the WhatsApp integration.

channels.whatsapp.bridge_url

string

default:"ws://localhost:3001"

WebSocket URL of the WhatsApp bridge server.

channels.whatsapp.use_native

boolean

default:"false"

Use native WhatsApp mode instead of a bridge.

channels.whatsapp.session_store_path

string

Path to store WhatsApp session data.

channels.whatsapp.allow_from

string[]

default:"[]"

Allowlist of WhatsApp phone numbers or JIDs.

channels.whatsapp.reasoning_channel_id

string

JID or chat ID for reasoning output.

`channels.feishu`

channels.feishu.enabled

boolean

default:"false"

Enable the Feishu/Lark bot.

channels.feishu.app_id

string

Feishu app ID.

channels.feishu.app_secret

string

Feishu app secret.

channels.feishu.encrypt_key

string

Feishu event encryption key.

channels.feishu.verification_token

string

Feishu webhook verification token.

channels.feishu.allow_from

string[]

default:"[]"

Allowlist of Feishu user open IDs.

channels.feishu.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.dingtalk`

channels.dingtalk.enabled

boolean

default:"false"

Enable the DingTalk bot.

channels.dingtalk.client_id

string

DingTalk app client ID.

channels.dingtalk.client_secret

string

DingTalk app client secret.

channels.dingtalk.allow_from

string[]

default:"[]"

Allowlist of DingTalk user IDs.

channels.dingtalk.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.qq`

channels.qq.enabled

boolean

default:"false"

Enable the QQ bot.

channels.qq.app_id

string

QQ app ID.

channels.qq.app_secret

string

QQ app secret.

channels.qq.allow_from

string[]

default:"[]"

Allowlist of QQ user IDs.

channels.qq.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.line`

channels.line.enabled

boolean

default:"false"

Enable the LINE bot.

channels.line.channel_secret

string

LINE channel secret.

channels.line.channel_access_token

string

LINE channel access token.

channels.line.webhook_path

string

default:"/webhook/line"

Webhook path for LINE events.

channels.line.allow_from

string[]

default:"[]"

Allowlist of LINE user IDs.

channels.line.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.onebot`

channels.onebot.enabled

boolean

default:"false"

Enable the OneBot v11 adapter (QQ via go-cqhttp or similar).

channels.onebot.ws_url

string

default:"ws://127.0.0.1:3001"

WebSocket URL of the OneBot v11 server.

channels.onebot.access_token

string

Access token for the OneBot server.

channels.onebot.reconnect_interval

number

default:"5"

Reconnection interval in seconds when the WebSocket connection drops.

channels.onebot.group_trigger

object

Show properties

prefixes

string[]

Respond only when a group message starts with one of these prefixes.

channels.onebot.allow_from

string[]

default:"[]"

Allowlist of QQ user IDs.

channels.onebot.reasoning_channel_id

string

Group or user ID for reasoning output.

`channels.maixcam`

channels.maixcam.enabled

boolean

default:"false"

Enable the MaixCam hardware channel.

channels.maixcam.host

string

default:"0.0.0.0"

Bind address for the MaixCam listener.

channels.maixcam.port

number

default:"18790"

Port for the MaixCam listener.

channels.maixcam.allow_from

string[]

default:"[]"

Allowlist of device IDs.

channels.maixcam.reasoning_channel_id

string

Channel ID for reasoning output.

`channels.wecom`

WeCom Bot — easier setup, supports group chats via webhook.

channels.wecom.enabled

boolean

default:"false"

Enable the WeCom bot channel.

channels.wecom.token

string

WeCom verification token.

channels.wecom.encoding_aes_key

string

43-character AES encoding key for message encryption.

channels.wecom.webhook_url

string

WeCom group robot webhook URL.

channels.wecom.webhook_path

string

default:"/webhook/wecom"

Incoming webhook path.

channels.wecom.allow_from

string[]

default:"[]"

Allowlist of WeCom user IDs.

channels.wecom.reply_timeout

number

default:"5"

Reply timeout in seconds.

channels.wecom.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.wecom_app`

WeCom self-built application — more features, proactive messaging, private chats only.

channels.wecom_app.enabled

boolean

default:"false"

Enable the WeCom App channel.

channels.wecom_app.corp_id

string

WeCom corporation ID.

channels.wecom_app.corp_secret

string

WeCom application secret.

channels.wecom_app.agent_id

number

WeCom application agent ID.

channels.wecom_app.token

string

WeCom verification token.

channels.wecom_app.encoding_aes_key

string

43-character AES encoding key.

channels.wecom_app.webhook_path

string

default:"/webhook/wecom-app"

Incoming webhook path.

channels.wecom_app.allow_from

string[]

default:"[]"

Allowlist of WeCom user IDs.

channels.wecom_app.reply_timeout

number

default:"5"

Reply timeout in seconds.

channels.wecom_app.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.wecom_aibot`

WeCom AI Bot (智能机器人) — official WeCom AI Bot integration with proactive messaging.

channels.wecom_aibot.enabled

boolean

default:"false"

Enable the WeCom AI Bot channel.

channels.wecom_aibot.token

string

WeCom verification token.

channels.wecom_aibot.encoding_aes_key

string

43-character AES encoding key.

channels.wecom_aibot.webhook_path

string

default:"/webhook/wecom-aibot"

Incoming webhook path.

channels.wecom_aibot.max_steps

number

default:"10"

Maximum streaming steps per response.

channels.wecom_aibot.welcome_message

string

Message sent automatically when a user opens a chat. Set to "" to disable.

channels.wecom_aibot.allow_from

string[]

default:"[]"

Allowlist of WeCom user IDs.

channels.wecom_aibot.reasoning_channel_id

string

Chat ID for reasoning output.

`channels.standard`

HTTP/WebSocket standard channel for custom integrations.

channels.standard.enabled

boolean

default:"false"

Enable the standard WebSocket channel.

channels.standard.token

string

Bearer token for authenticating incoming connections.

channels.standard.allow_origins

string[]

CORS allowed origins.

channels.standard.ping_interval

number

default:"30"

WebSocket ping interval in seconds.

channels.standard.read_timeout

number

default:"60"

Read timeout in seconds.

channels.standard.write_timeout

number

default:"10"

Write timeout in seconds.

channels.standard.max_connections

number

default:"100"

Maximum simultaneous WebSocket connections.

channels.standard.allow_from

string[]

default:"[]"

Allowlist of client IDs.

`tools`

tools.allow_read_paths

string[]

Additional filesystem paths the agent is permitted to read from, even when restrict_to_workspace is true.

tools.allow_write_paths

string[]

Additional filesystem paths the agent is permitted to write to, even when restrict_to_workspace is true.

`tools.web`

Controls web search and fetch capabilities.

tools.web.proxy

string

HTTP/HTTPS/SOCKS5 proxy URL for all web tool requests. For proxies that require credentials, use the HTTP_PROXY/HTTPS_PROXY environment variables instead of embedding credentials here.

tools.web.fetch_limit_bytes

number

default:"10485760"

Maximum number of bytes to fetch from a single URL. Default is 10 MB.

tools.web.brave

object

Brave Search integration.

Show properties

enabled

boolean

default:"false"

Enable Brave Search.

api_key

string

Brave Search API key. Obtain from the Brave Search API dashboard.

max_results

number

default:"5"

Maximum number of search results returned per query.

tools.web.duckduckgo

object

DuckDuckGo search (no API key required).

Show properties

enabled

boolean

default:"true"

Enable DuckDuckGo search.

max_results

number

default:"5"

Maximum number of search results returned per query.

tools.web.perplexity

object

Perplexity AI search integration.

Show properties

enabled

boolean

default:"false"

Enable Perplexity search.

api_key

string

Perplexity API key (pplx-…).

max_results

number

default:"5"

Maximum number of search results returned per query.

tools.web.tavily

object

Tavily search integration.

Show properties

enabled

boolean

default:"false"

Enable Tavily search.

api_key

string

Tavily API key.

base_url

string

Override the default Tavily API base URL.

max_results

number

default:"5"

Maximum number of results per query.

`tools.exec`

Controls shell command execution safety.

tools.exec.enable_deny_patterns

boolean

default:"true"

When true, a built-in set of dangerous command patterns is blocked before execution. Set to false to disable all built-in blocking (not recommended in production).Blocked by default: rm -rf, format, mkfs, dd if=, shutdown, reboot, pkill, killall, sudo, chmod, chown, curl | sh, wget | sh, docker run, docker exec, git push --force, eval, and others.

tools.exec.custom_deny_patterns

string[]

default:"[]"

Additional regular expressions to block. Commands matching any pattern are rejected. Patterns follow Go regexp syntax.

{
  "custom_deny_patterns": ["\\brm\\s+-r\\b", "\\bkillall\\s+python"]
}

tools.exec.custom_allow_patterns

string[]

default:"[]"

Regular expressions for commands that are explicitly allowed, overriding any matching deny patterns.

`tools.cron`

tools.cron.exec_timeout_minutes

number

default:"5"

Maximum execution time in minutes for cron-triggered commands. Set to 0 for no timeout.

`tools.mcp`

Model Context Protocol server integration. MCP extends the agent with tools from external servers.

tools.mcp.enabled

boolean

default:"false"

Global toggle for all MCP server connections. Must be true for any server to connect.

tools.mcp.servers

object

Map of server names to server configuration objects. Server names are used only for identification in logs.

Show per-server properties

enabled

boolean

required

Enable this specific MCP server.

type

string

Transport type: "stdio", "sse", or "http". Auto-detected when omitted: if url is set → sse; if command is set → stdio.

command

string

Executable to launch for stdio transport (e.g., "npx", "python").

args

string[]

Arguments passed to command for stdio transport.

env

object

Environment variables injected into the stdio process. Use this for API keys the MCP server needs.

env_file

string

Path to a .env-format file whose variables are loaded into the stdio process environment.

url

string

Endpoint URL for sse or http transport.

headers

object

HTTP headers sent with every request to sse/http servers. Commonly used for Authorization tokens.

Stdio example:

{
  "tools": {
    "mcp": {
      "enabled": true,
      "servers": {
        "filesystem": {
          "enabled": true,
          "command": "npx",
          "args": ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
        }
      }
    }
  }
}

Remote SSE/HTTP example:

{
  "tools": {
    "mcp": {
      "enabled": true,
      "servers": {
        "context7": {
          "enabled": true,
          "type": "http",
          "url": "https://mcp.context7.com/mcp",
          "headers": { "CONTEXT7_API_KEY": "ctx7sk-xx" }
        }
      }
    }
  }
}

MCP server configuration (tools.mcp.servers.*) cannot be set via environment variables. Use config.json for all server definitions.

`tools.skills`

Skill discovery and installation via registries.

tools.skills.registries.clawhub

object

ClawHub skill registry configuration.

Show properties

enabled

boolean

default:"true"

Enable the ClawHub registry for skill search and installation.

base_url

string

default:"https://clawhub.ai"

ClawHub API base URL.

auth_token

string

Optional authentication token for the registry.

search_path

string

default:"/api/v1/search"

Search endpoint path.

skills_path

string

default:"/api/v1/skills"

Skills list endpoint path.

download_path

string

default:"/api/v1/download"

Skill download endpoint path.

timeout

number

Request timeout in seconds for registry API calls.

max_zip_size

number

Maximum ZIP file size in bytes for skill downloads.

max_response_size

number

Maximum response size in bytes for registry API calls.

tools.skills.max_concurrent_searches

number

default:"2"

Maximum number of simultaneous registry search requests.

tools.skills.search_cache

object

In-memory search result cache settings.

Show properties

max_size

number

default:"50"

Maximum number of cached search results.

ttl_seconds

number

default:"300"

Cache TTL in seconds (default: 5 minutes).

`tools.media_cleanup`

Automatic cleanup of temporary media files.

tools.media_cleanup.enabled

boolean

default:"true"

Enable automatic media file cleanup.

tools.media_cleanup.max_age_minutes

number

default:"30"

Files older than this are eligible for removal.

tools.media_cleanup.interval_minutes

number

default:"5"

How often the cleanup job runs.

`heartbeat`

heartbeat.enabled

boolean

default:"true"

Send periodic heartbeat pings to the Operator cloud service for uptime monitoring and remote management. Disable if running in an air-gapped environment.

heartbeat.interval

number

default:"30"

Heartbeat interval in minutes. Minimum effective value is 5.

`devices`

devices.enabled

boolean

default:"false"

Enable device monitoring capabilities (currently USB device detection).

devices.monitor_usb

boolean

default:"true"

Monitor USB device plug/unplug events when devices.enabled is true.

`gateway`

The local HTTP/WebSocket gateway that exposes the Operator API on the local machine.

gateway.host

string

default:"127.0.0.1"

Bind address for the gateway listener. Change to "0.0.0.0" to accept connections from the network.

gateway.port

number

default:"18790"

TCP port for the gateway listener.

Deprecated: `providers`

The top-level providers key is deprecated as of v1.x and will be removed in v2.0. Migrate to model_list instead. See the migration guide for details.If model_list is empty and providers is present, Operator performs an automatic in-memory migration at startup.

The legacy format configured providers globally rather than per-model:

{
  "providers": {
    "openai": { "api_key": "sk-...", "api_base": "https://api.openai.com/v1" },
    "anthropic": { "api_key": "sk-ant-..." }
  },
  "agents": {
    "defaults": { "provider": "openai", "model": "gpt-5.2" }
  }
}

CLI Commands

Configuration Reference

Configuration file reference

`agents`

`model_list`

Load balancing

`channels`

`channels.telegram`

`channels.discord`

`channels.slack`

`channels.whatsapp`

`channels.feishu`

`channels.dingtalk`

`channels.qq`

`channels.line`

`channels.onebot`

`channels.maixcam`

`channels.wecom`

`channels.wecom_app`

`channels.wecom_aibot`

`channels.standard`

`tools`

`tools.web`

`tools.exec`

`tools.cron`

`tools.mcp`

`tools.skills`

`tools.media_cleanup`

`heartbeat`

`devices`

`gateway`

Deprecated: `providers`

Build docs developers (and LLMs) love

CLI Commands

Configuration Reference

​agents

​model_list

​Load balancing

​channels

​channels.telegram

​channels.discord

​channels.slack

​channels.whatsapp

​channels.feishu

​channels.dingtalk

​channels.qq

​channels.line

​channels.onebot

​channels.maixcam

​channels.wecom

​channels.wecom_app

​channels.wecom_aibot

​channels.standard

​tools

​tools.web

​tools.exec

​tools.cron

​tools.mcp

​tools.skills

​tools.media_cleanup

​heartbeat

​devices

​gateway

​Deprecated: providers

Build docs developers (and LLMs) love

`agents`

`model_list`

Load balancing

`channels`

`channels.telegram`

`channels.discord`

`channels.slack`

`channels.whatsapp`

`channels.feishu`

`channels.dingtalk`

`channels.qq`

`channels.line`

`channels.onebot`

`channels.maixcam`

`channels.wecom`

`channels.wecom_app`

`channels.wecom_aibot`

`channels.standard`

`tools`

`tools.web`

`tools.exec`

`tools.cron`

`tools.mcp`

`tools.skills`

`tools.media_cleanup`

`heartbeat`

`devices`

`gateway`

Deprecated: `providers`