### Run MetaClaw Setup and Start

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/README.md

Commands to run after installing the MetaClaw plugin to set it up and start the service.

```bash
metaclaw setup
metaclaw start
```

--------------------------------

### MetaClaw Setup and Start Commands

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

These commands are used for initial configuration and launching the MetaClaw agent. The 'start' command can be customized with different modes.

```bash
metaclaw setup              # one-time config wizard
metaclaw start              # default: auto mode — skills + scheduled RL training
metaclaw start --mode rl    # RL without scheduler (trains immediately on full batch)
metaclaw start --mode skills_only  # skills only, no RL (no Tinker needed)
```

--------------------------------

### GET /users Response Example (200 OK)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example JSON response for a successful GET /users request, including user data and pagination details.

```json
{
  "data": [
    {
      "id": "usr_abc123",
      "email": "user@example.com",
      "name": "John Doe",
      "role": "developer",
      "created_at": "2026-03-15T10:30:00Z"
    }
  ],
  "pagination": {
    "page": 1,
    "limit": 20,
    "total": 45
  }
}
```

--------------------------------

### GET /users Request Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example of a GET request to retrieve a list of users, with pagination parameters.

```http
GET /users?page=1&limit=20
```

--------------------------------

### GET /tasks Response Example (200 OK)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example JSON response for a successful GET /tasks request, including task data and pagination details.

```json
{
  "data": [
    {
      "id": "tsk_001",
      "title": "Implement authentication",
      "description": "Add JWT-based authentication to the API",
      "status": "active",
      "assignee": "usr_abc123",
      "created_at": "2026-04-01T14:20:00Z",
      "due_date": "2026-04-10T17:00:00Z"
    }
  ],
  "pagination": {
    "page": 1,
    "limit": 20,
    "total": 8
  }
}
```

--------------------------------

### Setup MetaClaw Environment Variables

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/docs/scripts.md

Copy the example environment file and source it before running scripts. Ensure you edit the copied file with your specific values.

```bash
cp benchmark/scripts/_env_arg_example.sh ~/.metaclaw_bench_env.sh
# Edit ~/.metaclaw_bench_env.sh with your values
source ~/.metaclaw_bench_env.sh
```

--------------------------------

### Run MetaClaw setup

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Initiates the interactive setup wizard for MetaClaw configuration.

```bash
metaclaw setup
```

--------------------------------

### Structured Logging Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/memory_data/skills/structured-logging-and-observability/SKILL.md

Demonstrates how to use structlog to emit structured JSON logs with contextual information. Ensure structlog is installed and configured.

```python
import structlog
log = structlog.get_logger()
log.info("request_complete", method="POST", path="/api/data", status=200, latency_ms=42)
```

--------------------------------

### GET /tasks Response Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day27/api_reference.md

Example JSON response when retrieving a list of tasks. Includes task details and total count.

```json
{
  "tasks": [
    {
      "id": "task_001",
      "title": "Implement user authentication",
      "status": "in_progress",
      "assignee": "user_42",
      "created": "2026-03-15",
      "due": "2026-04-10"
    }
  ],
  "total": 1
}
```

--------------------------------

### GET /tasks Request Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example of a GET request to retrieve a list of tasks, with filtering parameters for status and assignee.

```http
GET /tasks?status=active&assignee=usr_abc123
```

--------------------------------

### Start Development Server

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day11/existing_readme.md

Start the local development server. The application will be accessible at http://localhost:3000.

```bash
npm run dev
```

--------------------------------

### Memory Mode Experiment Setup and Run

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Configures LLM credentials and project root, then executes the memory mode script. This script automatically starts a proxy for memory-only mode.

```bash
export BENCHMARK_BASE_URL=https://api.openai.com/v1
export BENCHMARK_API_KEY=sk-...
export BENCHMARK_MODEL=gpt-5.2
export METACLAW_ROOT=/absolute/path/to/MetaClaw

# memory_run.py starts a proxy — no extra vars needed for memory-only mode
python /absolute/path/to/MetaClaw/benchmark/scripts/memory_run.py
```

--------------------------------

### Copy Environment Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day11/existing_readme.md

Copy the example environment file and edit it with your specific configuration settings.

```bash
cp .env.example .env
# Edit .env with your configuration
```

--------------------------------

### Install Project Dependencies

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day11/existing_readme.md

Install all necessary project dependencies using npm.

```bash
npm install
```

--------------------------------

### Install OpenClaw with curl

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs the OpenClaw plugin using curl, unzips it, enables the plugin, and restarts the gateway. This is a one-click installation method.

```bash
curl -LO https://github.com/aiming-lab/MetaClaw/releases/download/v0.4.0/metaclaw-plugin.zip
unzip metaclaw-plugin.zip -d ~/.openclaw/extensions
openclaw plugins enable metaclaw-openclaw && openclaw gateway restart
```

--------------------------------

### POST /users Request Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example of a POST request to create a new user, including the request body with user details.

```http
POST /users
Content-Type: application/json

{
  "email": "newuser@example.com",
  "name": "Jane Smith",
  "role": "developer"
}
```

--------------------------------

### Install Metaclaw Memory Plugin

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/README.md

Install the Metaclaw Memory plugin using the OpenClaw CLI. This also includes setting up the Python sidecar and verifying the installation.

```bash
openclaw plugins install @metaclaw/memory

openclaw metaclaw setup

openclaw metaclaw status
```

--------------------------------

### Install MetaClaw Memory Plugin

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/docs/QUICK_START.md

Install the plugin from npm or a local development directory.

```bash
# From npm (after publish)
openclaw plugins install @metaclaw/memory

# Or from local directory (dev)
openclaw plugins install -l /path/to/openclaw-metaclaw-memory
```

--------------------------------

### POST /users Response Example (201 Created)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example JSON response for a successful POST /users request, showing the newly created user's data.

```json
{
  "data": {
    "id": "usr_xyz789",
    "email": "newuser@example.com",
    "name": "Jane Smith",
    "role": "developer",
    "created_at": "2026-04-06T09:15:00Z"
  }
}
```

--------------------------------

### Registering CLI Commands with OpenClaw

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/PLAN.md

Example of registering CLI commands using `api.registerCli`, which allows for custom commands like `openclaw metaclaw setup`.

```typescript
import { program } from "commander";
import setupCommand from "./commands/setup";
import statusCommand from "./commands/status";
import searchCommand from "./commands/search";
import wipeCommand from "./commands/wipe";
import upgradeCommand from "./commands/upgrade";

export const register = (api) => {
  const metaclaw = program.command("metaclaw");
  setupCommand(metaclaw);
  statusCommand(metaclaw);
  searchCommand(metaclaw);
  wipeCommand(metaclaw);
  upgradeCommand(metaclaw);

  api.registerCli({ program });
};
```

--------------------------------

### Install OpenClaw Plugin Locally

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/OPENCLAW_PLUGIN_SPEC.md

Install the plugin from a local directory for development or testing purposes. Ensure you have run npm install and built the project first.

```bash
cd /path/to/openclaw-metaclaw-memory
npm install && npm run build
openclaw plugins install -l .
```

--------------------------------

### Install MetaClaw with pip (skills_only)

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs MetaClaw in skills_only mode using pip, which is a lightweight installation.

```bash
pip install -e .
```

--------------------------------

### Install WeChat plugin manually

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Manually installs the WeChat plugin using npx. This is an alternative to automatic installation.

```bash
npx -y @tencent-weixin/openclaw-weixin-cli@latest install
```

--------------------------------

### Install OpenClaw Plugin from Tarball

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/OPENCLAW_PLUGIN_SPEC.md

Install the plugin by first creating a tarball using npm pack, then installing the generated .tgz file.

```bash
npm pack
openclaw plugins install ./metaclaw-memory-0.1.0.tgz
```

--------------------------------

### Install and Run MetaClaw Benchmark

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Installs the MetaClaw benchmark, sets up required environment variables, validates the dataset, and runs the full pipeline. Ensure Python version is 3.10 or higher.

```bash
# Install (requires Python ≥ 3.10)
cd benchmark
pip install -e .

# Set required environment variables (see docs/scripts.md for the full list)
export BENCHMARK_BASE_URL=http://127.0.0.1:30000/v1   # OpenAI-compatible API base URL
export BENCHMARK_API_KEY=<your-api-key>                # API key for the above endpoint
export BENCHMARK_MODEL=<your-model-id>                 # Model ID expected by the API server

# Validate the dataset (run from the project root: MetaClaw/)
metaclaw-bench check -p benchmark/data/metaclaw-bench/all_tests.json

# Run the full pipeline (infer → scoring → report)
metaclaw-bench run \
  -i benchmark/data/metaclaw-bench/all_tests.json \
  -o benchmark/results
```

--------------------------------

### Install OpenClaw Plugin from npm

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/OPENCLAW_PLUGIN_SPEC.md

Use this command to install the plugin from the npm registry after it has been published.

```bash
openclaw plugins install @metaclaw/memory
```

--------------------------------

### POST /tasks Response Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day27/api_reference.md

Example JSON response after successfully creating a new task. Includes the new task's ID and status.

```json
{
  "id": "task_002",
  "title": "Task title",
  "status": "open",
  "created": "2026-04-10"
}
```

--------------------------------

### List Installed OpenClaw Plugins

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/OPENCLAW_PLUGIN_SPEC.md

Verify that the metaclaw-memory plugin has been successfully installed and is discoverable by OpenClaw.

```bash
openclaw plugins list
# metaclaw-memory should appear
```

--------------------------------

### Install MetaClaw with pip (full RL + scheduler)

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs MetaClaw with recommended full support for RL training and scheduler integration.

```bash
pip install -e ".[rl,evolve,scheduler]"
```

--------------------------------

### Install MetaClaw with pip (skill evolution)

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs MetaClaw with support for skill evolution using an OpenAI-compatible LLM.

```bash
pip install -e ".[evolve]"
```

--------------------------------

### Initialize MetaClaw Memory

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/docs/QUICK_START.md

Set up the Python virtual environment and install the sidecar for MetaClaw.

```bash
openclaw metaclaw setup    # creates Python venv + installs sidecar
openclaw metaclaw status   # verify: should show "Status: ok"
```

--------------------------------

### Install MetaClaw with pip (RL training support)

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs MetaClaw with support for RL training, including torch, transformers, and tinker.

```bash
pip install -e ".[rl]"
```

--------------------------------

### Start MetaClaw service

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Starts the MetaClaw service, which includes the proxy, agent configuration, and gateway restart.

```bash
metaclaw start
```

--------------------------------

### Install MetaClaw with pip (scheduler integration)

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Installs MetaClaw with Google Calendar integration for the scheduler.

```bash
pip install -e ".[scheduler]"
```

--------------------------------

### MetaClaw CLI Commands

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Commonly used MetaClaw CLI commands for setup, operation, and configuration management. Use `metaclaw setup` for initial configuration.

```bash
metaclaw setup                  # Interactive first-time configuration wizard
```

```bash
metaclaw start                  # Start MetaClaw (default: auto mode)
```

```bash
metaclaw start --mode rl        # Force RL mode (no scheduler) for this session
```

```bash
metaclaw start --mode skills_only  # Force skills-only mode for this session
```

```bash
metaclaw stop                   # Stop a running MetaClaw instance
```

```bash
metaclaw status                 # Check proxy health, running mode, and scheduler state
```

```bash
metaclaw config show            # View current configuration
```

```bash
metaclaw config KEY VALUE       # Set a config value
```

```bash
metaclaw config llm.oauth_token TOKEN        # Store OAuth token for current CLI provider
```

```bash
metaclaw auth paste-token --provider anthropic      # Store OAuth token (anthropic | openai-codex | gemini)
```

```bash
metaclaw auth status                                # Show all stored auth profiles
```

```bash
metaclaw uninstall              # Remove all MetaClaw data, OpenClaw extension, and pip package
```

--------------------------------

### POST /tasks Request Body Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day27/api_reference.md

Example JSON request body for creating a new task. Specify title, description, assignee, and due date.

```json
{
  "title": "Task title",
  "description": "Task description",
  "assignee": "user_42",
  "due": "2026-04-15"
}
```

--------------------------------

### PUT /tasks/{task_id} Response Example (200 OK)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day16/api_docs.md

Example JSON response for a successful PUT /tasks/{task_id} request, showing the updated task data.

```json
{
  "data": {
    "id": "tsk_001",
    "title": "Implement authentication",
    "status": "completed",
    "notes": "Authentication implementation completed and tested",
    "updated_at": "2026-04-06T11:30:00Z"
  }
}
```

--------------------------------

### Programmatic Rollout Task Example

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

An example JSON object representing a task for programmatic rollout in MetaClaw. This format is used when setting the 'openclaw_env_data_dir' environment variable.

```json
{"task_id": "task_1", "instruction": "Register the webhook at https://example.com/hook"}
```

--------------------------------

### Updating openclaw.plugin.json

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/PLAN.md

Example of a standard openclaw.plugin.json manifest file, including id, kind, and configSchema.

```json
{
  "id": "metaclaw-memory",
  "name": "MetaClaw Memory",
  "kind": "memory",
  "configSchema": { ... },
  "uiHints": { ... }
}
```

--------------------------------

### Full Auto OpenCLAW Config

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/metaclaw-openclaw/README.md

Enables the metaclaw-openclaw plugin with one-click MetaClaw setup. This option handles venv creation, pip installation, and starting MetaClaw automatically.

```json
{
  "plugins": {
    "entries": {
      "metaclaw-openclaw": {
        "enabled": true,
        "config": {
          "oneClickMetaclaw": true
        }
      }
    }
  }
}
```

--------------------------------

### Initialize Database

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/data/metaclaw-bench/workspaces/shared/day11/existing_readme.md

Run database migrations and seeding to set up the initial database state.

```bash
npm run db:migrate
npm run db:seed
```

--------------------------------

### Baseline Experiment Setup and Run

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Sets up LLM credentials and project root, then runs the baseline experiment script without a proxy. This is for basic model evaluation.

```bash
# 1. Required LLM credentials
export BENCHMARK_BASE_URL=https://api.openai.com/v1
export BENCHMARK_API_KEY=sk-...
export BENCHMARK_MODEL=gpt-5.2           # model ID expected by your API server

# 2. Strongly recommended: pin the project root to an absolute path
export METACLAW_ROOT=/absolute/path/to/MetaClaw

# 3. Run
python /absolute/path/to/MetaClaw/benchmark/scripts/baseline_run.py
```

--------------------------------

### Skills Mode Experiment Setup and Run

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Sets up LLM credentials, project root, and skills directory, then runs the skills-only experiment script. Ensure METACLAW_SKILLS_DIR is correctly set.

```bash
export BENCHMARK_BASE_URL=https://api.openai.com/v1
export BENCHMARK_API_KEY=sk-...
export BENCHMARK_MODEL=gpt-5.2
export METACLAW_ROOT=/absolute/path/to/MetaClaw
export METACLAW_SKILLS_DIR=/absolute/path/to/your/skills   # required for skills scripts

python /absolute/path/to/MetaClaw/benchmark/scripts/skills_only_run.py
```

--------------------------------

### Minimal OpenCLAW Config

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/metaclaw-openclaw/README.md

Enables the metaclaw-openclaw plugin. Requires manual setup and start commands.

```json
{
  "plugins": {
    "entries": {
      "metaclaw-openclaw": {
        "enabled": true
      }
    }
  }
}
```

--------------------------------

### Enable WeChat integration

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Enables WeChat integration by configuring MetaClaw and starting the service. The plugin is installed automatically.

```bash
metaclaw config wechat.enabled true
metaclaw start
```

--------------------------------

### RL Mode Experiment Setup and Run

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Configures environment variables for LLM, project root, and RL training, including Tinker and reward model IDs. Then, it runs the RL-only experiment script.

```bash
# Skill evolver uses BENCHMARK_* vars
# export BENCHMARK_BASE_URL=https://api.openai.com/v1
# export BENCHMARK_API_KEY=sk-...
# export BENCHMARK_MODEL=gpt-5.2
# export METACLAW_ROOT=/absolute/path/to/MetaClaw

# Additional vars for RL training
export TINKER_KEY=<tinker-api-key>
export TINKER_MODEL=<model-id-for-rl>
export PRM_MODEL=<process-reward-model-id>

python /absolute/path/to/MetaClaw/benchmark/scripts/rl_only_run.py
```

--------------------------------

### Configure MinT backend for RL

Source: https://github.com/aiming-lab/metaclaw/blob/main/README.md

Sets the RL backend to 'mint', configures the API key, base URL, and model for MinT.

```bash
metaclaw config rl.backend mint
metaclaw config rl.api_key sk-mint-...
metaclaw config rl.base_url https://mint.macaron.xin/
metaclaw config rl.model Qwen/Qwen3-4B-Instruct-2507
```

--------------------------------

### Run Benchmark with RL and Skills

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/docs/scripts.md

Executes the benchmark through the proxy with both RL training and pre-built skills active. Requires METACLAW_SKILLS_DIR and RL variables.

```bash
python benchmark/scripts/rl_run.py
```

--------------------------------

### Run Benchmark with RL and Memory (No Skills)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/docs/scripts.md

Combines RL training and memory features for the benchmark run, excluding skills. Requires RL variables.

```bash
python benchmark/scripts/rl_only_memory_run.py
```

--------------------------------

### Run Benchmark with All Features (MadMax)

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/docs/scripts.md

Enables all features simultaneously: RL training, skills, and memory. Requires METACLAW_SKILLS_DIR and RL variables.

```bash
python benchmark/scripts/madmax_memory_run.py
```

--------------------------------

### Create Swap File on Low RAM Systems

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/README.md

Command to create a 4GB swap file and enable it for systems with 2GB RAM or less, recommended before installing MetaClaw.

```bash
sudo fallocate -l 4G /swapfile && sudo chmod 600 /swapfile && sudo mkswap /swapfile && sudo swapon /swapfile
```

--------------------------------

### Install MetaClaw Plugin on macOS/Linux

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/README.md

Use this command to download and install the MetaClaw plugin on macOS and Linux systems.

```bash
curl -LO https://github.com/aiming-lab/MetaClaw/releases/download/v0.4.0/metaclaw-plugin.zip
unzip metaclaw-plugin.zip -d ~/.openclaw/extensions
openclaw plugins enable metaclaw-openclaw
openclaw gateway restart
```

--------------------------------

### Install MetaClaw Plugin on Windows (PowerShell)

Source: https://github.com/aiming-lab/metaclaw/blob/main/extensions/README.md

Use this PowerShell command to download and install the MetaClaw plugin on Windows systems.

```powershell
Invoke-WebRequest -Uri https://github.com/aiming-lab/MetaClaw/releases/download/v0.4.0/metaclaw-plugin.zip -OutFile metaclaw-plugin.zip
Expand-Archive metaclaw-plugin.zip -DestinationPath $env:USERPROFILE\.openclaw\extensions
openclaw plugins enable metaclaw-openclaw
openclaw gateway restart
```

--------------------------------

### Injecting Context via before_agent_start Hook

Source: https://github.com/aiming-lab/metaclaw/blob/main/openclaw-metaclaw-memory/PLAN.md

How to return context to be prepended to the agent's prompt before the agent starts execution.

```javascript
return { prependContext: "..." }
```

--------------------------------

### Install Development Dependencies and Run Tests

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/README.md

Installs Metaclaw in editable mode with development dependencies and runs all tests using pytest.

```bash
pip install -e ".[dev]"
pytest -v tests/
```

--------------------------------

### Run Benchmark with Skills Only

Source: https://github.com/aiming-lab/metaclaw/blob/main/benchmark/docs/scripts.md

Runs the benchmark through the proxy with pre-built skills injected into the agent context, excluding memory and RL. Requires METACLAW_SKILLS_DIR.

```bash
python benchmark/scripts/skills_only_run.py
```