---
title: OpenWorker
description: How ClawMetry observes OpenWorker — where its data lives, the environment variables that move it, what cost data is real, and the traps in its store.
eyebrow: Runtime reference
---

# OpenWorker

OpenWorker is a local-first desktop agent that drives 25-plus SaaS connectors alongside the shell and filesystem. Its stated output is finished work rather than a chat transcript, which makes it the least coding-shaped runtime in the matrix.

| | |
|---|---|
| **Runtime id** | `openworker` |
| **Category** | Generalist desktop worker |
| **Vendor** | Andrew Ng |
| **Tier** | Paid tier (Starter and up) |
| **Upstream** | [github.com/andrewyng/openworker](https://github.com/andrewyng/openworker) |

## Where the data comes from

ClawMetry reads these locations, in order, and never writes to any of them.

- `$COWORKER_STATE_DIR/coworker.db` — sessions, workspaces and audit events
- `$COWORKER_STATE_DIR/conversations/<sid>.jsonl` — append-only message log
- (default %APPDATA%\coworker on Windows, else ~/.config/coworker)

**Format.** SQLite index plus one JSONL message log per session; assistant messages carry `usage` and `reasoning` sidecars.

## Environment variables

| Variable | Effect |
|---|---|
| `COWORKER_STATE_DIR` | OpenWorker's own state dir |
| `CLAWMETRY_OPENWORKER_STATE_DIR` | ClawMetry-side override |

See [Environment variables](/docs/config/environment/) for the variables that apply to every runtime.

## Tokens and cost

**Tokens.** From the per-message `usage` sidecar, tagged with the model that produced it — so a session that switched models is priced per model and summed.

**Cost.** Always derived; OpenWorker writes no dollars anywhere.

:::note Why this section exists
Runtimes differ enormously in what they persist. Rather than showing a plausible number everywhere, ClawMetry reports what the store actually contains and marks the rest unknown. A cost of `$0.00` and a cost of *unknown* mean different things, and the dashboard keeps them different.
:::

## Things worth knowing

- The reviewer-token trap: `audit_events` also has token columns, but they meter the Auto-Approve *reviewer*, a separate and off-by-default LLM. Summing them into session cost reports a number that is both wrong and, by default, roughly zero. ClawMetry takes tokens only from the message sidecar and surfaces audit rows as approval events.
- `coworker.db` holds the session index and the audit log in the same file as two unrelated tables.

## Verify it is being read

```bash
# Is this runtime detected, and would this install unlock it?
clawmetry runtimes --json | jq '.runtimes[] | select(.id=="openworker")'

# Why is it locked, if it is?
clawmetry runtimes --why openworker

# Any sessions ingested yet?
curl -s 'http://localhost:8900/api/local/sessions?limit=5' \
  | jq '.rows[] | {session_id, agent_id, events, cost_usd}'
```

If detection succeeds but no sessions appear, work through [Troubleshooting](/docs/ops/troubleshooting/) — the usual causes are a relocated data directory (set the override above), a session limit that is hiding older runs, or a runtime that simply has not written anything yet.