FREE TOOL

OpenClaw Config Generator

Build openclaw.json visually.
Supports every field from the official schema. Export valid JSON5.

📖 Full Schema Reference

✓ Live JSON5 Preview ✓ Auto-save to browser

Identity

name, emoji, personality theme

Name

Emoji

Theme

Agent Defaults

workspace, model, skills, thinking, sandbox

Workspace Directory

Primary Model

Fallback Models

Thinking Level

Reasoning

Elevated

Verbose

Channels

telegram, discord, slack

Discord

Slack

Tools & Permissions

profile, allow, deny, elevated, exec

Base Profile

Minimal — session_status only Coding — fs, runtime, web, sessions, memory + cron, image, image_generate, video_generate Messaging — group:messaging, sessions_list, sessions_history, sessions_send, session_status Full Access — no restrictions

Allowed (extends profile)

Denied (blocks profile)

Elevated Exec

Allow exec outside sandbox — allowFrom set to all channels

Exec Settings

Background Timeout (ms)

Hard Timeout (sec)

Notify on process exit

group:fsread, write, edit, apply_patch

group:runtimeexec, process, code_execution

group:webweb_search, x_search, web_fetch

group:sessionssessions_list, sessions_history, sessions_send, sessions_spawn, sessions_yield, subagents, session_status

group:memorymemory_search, memory_get

group:uibrowser, canvas

group:automationcron, gateway

group:messagingmessage

group:mediaimage, image_generate, video_generate, tts

group:nodesnodes

group:agentsagents_list

group:openclawall built-in tools (excludes provider plugins)

Tools not in any group (allow/deny individually)

update_plan pdf pdfs dir_fetch dir_list file_fetch file_write wiki_search wiki_get wiki_apply wiki_lint wiki_status music_generate

Models & Providers

custom Ollama, LM Studio, LiteLLM, etc.

Catalog Mode

Add API keys for OpenRouter, OpenAI,

Anthropic, Groq, DeepSeek & 50+ more.

Custom Providers

No custom providers added yet.
Built-in providers (OpenRouter, OpenAI, etc.) need only an API key.
Self-hosted (Ollama, LM Studio) also need a base URL and model list.

openclaw.json

LIVE PREVIEW • UPDATES AUTOMATICALLY

This tool generates a complete openclaw.json file based on the official OpenClaw schema.

All changes are saved to browser localStorage automatically.

Built as a free public tool for the ClawVPS community.

Choose an Emoji

Guide & Reference

Understanding OpenClaw Configuration

Everything you need to know to configure your AI assistant the right way.

§ What is openclaw.json?

The openclaw.json file is the central configuration file for OpenClaw — the AI gateway that routes requests to language models, manages tools, channels, and agent behavior. When OpenClaw starts, it reads this file to understand how to behave: which model to use, what tools are available, which messaging platforms to connect to, and how to handle user requests.

This visual config generator lets you build a valid openclaw.json without memorizing the schema — every field is documented, typed, and validated in real time.

§ Identity — Give Your Agent a Personality

The Identity section defines how your AI assistant presents itself. The name field sets the agent's display name in logs and system messages. The emoji gives it a visual icon for interfaces that support it. The theme is a free-text personality descriptor that helps set the tone for how the agent communicates.

A good identity isn't just cosmetic — it influences how users perceive and interact with your assistant. A theme like "helpful coding assistant" vs "friendly customer support agent" shapes expectations even before a single word is exchanged.

§ Agent Defaults — Behavior Out of the Box

The Agent Defaults block controls how the agent behaves by default unless overridden per conversation. workspace sets the working directory for file operations and project context. primary and fallbacks define the model chain — OpenClaw will try the primary model first, then fall through to each fallback in order if the primary is unavailable or rate-limited.

The thinking level controls how much reasoning the model does before responding. Medium is the default for general use. High enables extended reasoning for complex tasks. Elevated determines whether the agent can execute system-level commands by default (set to ask to require user confirmation per session).

§ Channels — Connect Messaging Platforms

Channels are how your AI agent talks to users. OpenClaw supports WhatsApp, Telegram, and Discord out of the box. Each channel has its own configuration: bot tokens, user allowlists, and DM policies.

The DM Policy setting controls who can DM the bot: pairing (default) requires users to pair via QR code, allowlist restricts access to specific phone numbers or user IDs, and open allows anyone to interact. Use pairing for personal use, allowlist for team deployments, and open only when you intentionally want public access.

§ Tools & Permissions — What the Agent Can Do

The Tools section is the most security-sensitive part of your config. It controls which capabilities the agent has: reading/writing files, running shell commands, browsing the web, sending messages, and more.

The profile setting provides sensible presets: Minimal (session status only), Coding (file system + web + sessions), Messaging (group chat tools), or Full Access (no restrictions). You can extend or restrict any profile using the Allowed and Denied fields. For example, a Coding profile with deny: browser would remove browser control while keeping file operations.

Elevated Exec is a special flag that allows the agent to run commands outside the sandbox. When enabled, the agent can execute arbitrary shell commands with the permissions of the OpenClaw process. Only enable this if you fully trust the model and the inputs it receives — never on a publicly accessible instance.

§ Models & Providers — Powering the Agent

OpenClaw uses a provider-based model system. Built-in providers (OpenRouter, OpenAI, Anthropic, Google, Groq, and 50+ others) have their model catalogs pre-configured — you only need to provide your API key. OpenClaw handles model discovery and routing automatically.

For self-hosted models (Ollama, LM Studio, vLLM, SGLang), you need to provide the base URL of your local server and optionally list the available models. Since self-hosted servers aren't in the built-in catalog, OpenClaw can't auto-detect what models they expose without an explicit list.

The mode: merge setting (the default) merges your custom providers with OpenClaw's built-in catalog. Use mode: replace if you want to use only your own providers and exclude the built-in ones.

§ Exporting Your Configuration

When you're done, click DOWNLOAD to save your config as openclaw.json, or COPY to copy it to your clipboard. Place the file at ~/.openclaw/openclaw.json (or the path specified by OPENCLAW_CONFIG env var) before starting OpenClaw.

Your config is also auto-saved to browser localStorage, so you can return and continue editing without losing your work. Use Reset All to start fresh.

⚠️ Common Configuration Mistakes

✗ Leaving elevated exec enabled on a public-facing instance — a prompt injection attack could run arbitrary shell commands on your server. Always use elevated: ask or elevated: off for anything exposed to the internet.
✗ No fallback models — if your primary model goes down or hits rate limits, the agent becomes completely unavailable. Always specify at least one fallback in a different provider family.
✗ Open DM policy without allowlist — leaving dmPolicy: open with no restrictions means anyone who finds your bot token can interact with it.
✗ Putting sensitive keys in the wrong fields — API keys go in channel-specific fields, not in provider configs unless that provider actually needs them.

For the full configuration schema and all available options, see the official OpenClaw configuration reference.