> ## Documentation Index > Fetch the complete documentation index at: https://docs.getunbound.ai/llms.txt > Use this file to discover all available pages before exploring further. # OpenClaw > Enforce tool policies and route LLM calls through the Unbound AI Gateway for OpenClaw agents ## What is OpenClaw? OpenClaw is a self-hosted AI agent framework that runs on your hardware. It supports multi-agent workflows, tool execution, and integrations with MCP servers — all controlled through a local gateway. Unbound offers two independent integrations for OpenClaw: 1. **Tool Policies Plugin** — Enforce your organization's security policies on every tool call your OpenClaw agents make 2. **Unbound as LLM Provider** — Route OpenClaw's LLM calls through the Unbound gateway for visibility, guardrails, and model routing You can use either or both depending on your needs. ## Prerequisites * **OpenClaw**: Version 2026.2.0 or higher * **Node.js / npm**: Required for plugin installation * **Unbound Account**: Sign up at [gateway.getunbound.ai](https://gateway.getunbound.ai) if you don't have one *** ## Quick Setup (Recommended) The setup script handles everything — authentication, plugin installation, configuration, and LLM provider setup. **Open your terminal** and run: ```bash theme={null} python3 -c "$(curl -s https://getunbound.ai/setup/openclaw/install)" --domain gateway.getunbound.ai ``` This will: 1. Open your browser for authentication via the Unbound dashboard 2. Install the `unbound-openclaw-plugin` npm package globally 3. Configure the tool policies plugin in `~/.openclaw/openclaw.json` 4. Set up Unbound as the LLM provider 5. Set the `UNBOUND_OPENCLAW_API_KEY` environment variable in your shell profile Once complete, **restart your terminal** (or `source` your shell profile), then verify: ```bash theme={null} openclaw agent --local --agent main --message "hello world" ``` ### Setup Options You can choose which components to install: | Flag | Description | | ------------------ | -------------------------------------------------------- | | `--plugin` | Install only the tool policies plugin | | `--provider` | Install only the LLM provider | | `--model MODEL_ID` | Use a custom model (default: `claude-sonnet-4-20250514`) | If neither `--plugin` nor `--provider` is specified, both are installed. **Examples:** ```bash theme={null} # Plugin only (tool policies, no LLM routing) python3 -c "$(curl -s https://getunbound.ai/setup/openclaw/install)" --domain gateway.getunbound.ai --plugin # Provider only with a custom model python3 -c "$(curl -s https://getunbound.ai/setup/openclaw/install)" --domain gateway.getunbound.ai --provider --model fireworks-ai/kimi-k2p5 # Both with a custom model python3 -c "$(curl -s https://getunbound.ai/setup/openclaw/install)" --domain gateway.getunbound.ai --model claude-opus-4-20250514 ``` Available models can be found at [gateway.getunbound.ai/models](https://gateway.getunbound.ai/models). ### Removing the Integration To undo all changes (uninstalls plugin, removes config and environment variable): ```bash theme={null} python3 -c "$(curl -s https://getunbound.ai/setup/openclaw/install)" --clear ``` *** ## Verification After setup, test that tool policies are working: **Test 1: Safe command (should succeed)** ```bash theme={null} openclaw agent --local --agent main --message "run this shell command: echo hello world" ``` Expected: The command executes successfully. **Test 2: Blocked command (requires a Block policy for `delete_file`)** ```bash theme={null} openclaw agent --local --agent main --message "run this shell command: rm -rf /tmp/test.txt" ``` Expected: `[tools] exec failed: This command is blocked by your organization's policy.` ### Configuring Policies Tool policies are managed at [gateway.getunbound.ai/policies/tool-policies](https://gateway.getunbound.ai/policies/tool-policies). You can create policies for: * **Terminal commands** — e.g., blocking file deletions, database operations, or script execution * **MCP tools** — e.g., blocking Slack messages or GitHub issue creation Each policy can be set to **Block** (always denied) or **Warn** (allowed for OpenClaw, prompts for confirmation in Claude Code). *** ## Manual Configuration If you prefer to configure everything yourself instead of using the setup script: ### Tool Policies Plugin **Step 1:** Install the plugin: ```bash theme={null} npm install -g unbound-openclaw-plugin ``` **Step 2:** Add the plugin to `~/.openclaw/openclaw.json`: ```json theme={null} { "plugins": { "load": { "paths": ["/unbound-openclaw-plugin"] }, "entries": { "unbound-openclaw-plugin": { "enabled": true, "config": { "gatewayUrl": "https://api.getunbound.ai", "failOpen": true } } } } } ``` Run `npm root -g` to find your global npm modules path, then append `/unbound-openclaw-plugin` for the load path. **Step 3:** Set the API key in your shell profile: ```bash theme={null} export UNBOUND_OPENCLAW_API_KEY="your-api-key-here" ``` You can find your API key at [gateway.getunbound.ai/connect](https://gateway.getunbound.ai/connect). ### Plugin Configuration Reference | Option | Type | Default | Description | | ------------ | ------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | | `gatewayUrl` | string | *required* | Unbound gateway URL (`https://api.getunbound.ai`) | | `apiKey` | string | `$UNBOUND_OPENCLAW_API_KEY` | API key for authentication. Falls back to the environment variable | | `failOpen` | boolean | `true` | When `true`, tool calls proceed if the gateway is unreachable. Set to `false` to block all tool calls when the gateway is down | ### LLM Provider Add the provider to `~/.openclaw/openclaw.json`: ```json theme={null} { "agents": { "defaults": { "model": { "primary": "unbound/claude-sonnet-4-20250514" } } }, "models": { "providers": { "unbound": { "baseUrl": "https://api.getunbound.ai/v1", "apiKey": "${UNBOUND_OPENCLAW_API_KEY}", "api": "openai-completions", "models": [ { "id": "claude-sonnet-4-20250514", "name": "claude-sonnet-4-20250514", "contextWindow": 200000, "maxTokens": 8192 } ] } } } } ``` Replace `claude-sonnet-4-20250514` with any model available on the gateway. *** ## Security Benefits Using OpenClaw with Unbound Security AI Gateway provides: * **Request Monitoring**: All AI requests are logged and monitored * **Tool Policy Enforcement**: Block dangerous terminal commands and MCP tool calls * **Compliance**: Ensure AI interactions meet your organization's standards * **Audit Trail**: Complete visibility into AI usage patterns Configure terminal command and MCP tool policies Browse models available through the Unbound gateway