> ## Documentation Index
> Fetch the complete documentation index at: https://critiqor-71f5274a.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# critiqor monitor openclaw — Start Runtime Observation

> critiqor monitor openclaw launches OpenClaw and begins a Critiqor observation session. Key flags: --agent-id, --difficulty-tier, --timeout, --runs-dir.

`critiqor monitor openclaw` starts a Critiqor observation session and launches OpenClaw in the same terminal. Use this command whenever you want to capture a runtime evidence trace — whether you're validating a change, debugging a session, or running a benchmarked evaluation.

## Syntax

```bash theme={null}
critiqor monitor openclaw [OPTIONS]
```

## What It Does

When you run this command, Critiqor begins observing in the background and then launches OpenClaw in your terminal. The agent runs exactly as it normally would — you interact with it as usual while Critiqor captures runtime evidence without interfering with the session. When you're done, run `critiqor finalize` to complete the session and generate your reliability report.

## Options

| Flag                 | Type      | Default          | Description                                                                                       |
| -------------------- | --------- | ---------------- | ------------------------------------------------------------------------------------------------- |
| `--agent-id`         | string    | `openclaw_agent` | An identifier for this agent. Used to distinguish agents when you have multiple.                  |
| `--tenant-id`        | string    | `default`        | Tenant identifier for multi-tenant setups.                                                        |
| `--visibility`       | choice    | `private`        | Controls dashboard visibility. Choices: `private`, `public`, `anonymous`, `shared`.               |
| `--difficulty-tier`  | choice    | `standard`       | Sets the benchmark difficulty tier for evaluation. Choices: `easy`, `standard`, `hard`, `stress`. |
| `--cwd`              | directory | *(none)*         | Working directory to use when launching the agent.                                                |
| `--timeout`          | float     | *(none)*         | Optional time limit in seconds for the observation session.                                       |
| `--runs-dir`         | string    | `runs`           | Directory where run artifacts are stored.                                                         |
| `--openclaw-command` | string    | `openclaw chat`  | The OpenClaw command to launch.                                                                   |

## Examples

```bash theme={null}
# Start a standard observation session
critiqor monitor openclaw

# Use a custom agent ID and hard difficulty tier
critiqor monitor openclaw --agent-id my_agent --difficulty-tier hard

# Set a 5-minute timeout
critiqor monitor openclaw --timeout 300

# Launch from a specific working directory
critiqor monitor openclaw --cwd /path/to/project
```

## Expected Behavior

OpenClaw launches in the same terminal window immediately after observation begins. Work with it as you normally would — send messages, invoke tools, complete your task. Critiqor operates silently in the background throughout the session. When you're finished, run `critiqor finalize` to end the session and generate your reliability report.

## Common Use Cases

* **Validating a prompt change** — run a session before and after to compare reliability outcomes.
* **Testing a new tool** — observe how the agent behaves with a newly integrated tool under standard conditions.
* **Debugging a failed session** — reproduce the issue under observation to capture a detailed evidence trace.
* **Stress-testing** — use `--difficulty-tier stress` to evaluate agent behavior under high-demand conditions.