# lockn-logger

> Automatically captures LLM API call usage from OpenClaw sessions and logs them to LockN Logger for cost tracking and attribution.

- Author: LockN-AI
- Repository: LockN-AI/lockn-logger
- Version: 20260205185742
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/LockN-AI/lockn-logger
- Web: https://mule.run/skillshub/@@LockN-AI/lockn-logger~lockn-logger:20260205185742

---

# LockN Logger Skill

Automatically captures LLM API call usage from OpenClaw sessions and logs them to LockN Logger for cost tracking and attribution.

## Description

This skill monitors OpenClaw session transcripts and extracts LLM usage data (tokens, costs, model info) to post to a LockN Logger instance. Enables zero-friction cost tracking across all agents.

## Usage

The skill runs automatically via cron or can be triggered manually. It scans session transcripts for messages with `usage` data and posts receipts to the configured LockN Logger endpoint.

### Manual Run
```bash
node /path/to/skills/lockn-logger/capture.js
```

### Cron Integration
Add to HEARTBEAT.md or configure a cron job to run periodically.

## Configuration

Set these environment variables or configure in OpenClaw config:

| Variable | Default | Description |
|----------|---------|-------------|
| `LOCKN_LOGGER_ENDPOINT` | `http://localhost:8080/api/receipts` | LockN Logger API URL |
| `LOCKN_LOGGER_API_KEY` | (none) | Optional API key for auth |
| `LOCKN_LOGGER_BATCH_SIZE` | `10` | Receipts per batch POST |
| `LOCKN_LOGGER_INCLUDE_LOCAL` | `true` | Include local models (cost=$0) |

## What Gets Captured

From each OpenClaw message with usage data:
- **model**: Model identifier (e.g., `claude-opus-4-5`)
- **provider**: Provider name (e.g., `anthropic`, `llamacpp-qwen`)
- **inputTokens**: Input token count
- **outputTokens**: Output token count
- **totalCost**: Calculated cost in USD
- **timestamp**: When the call was made
- **correlationId**: Session ID for attribution
- **agentId**: Agent identifier from session key

## Deduplication

The skill tracks the last processed timestamp per session to avoid duplicate receipts. State is stored in `state/last-scan.json`.

## Error Handling

- Network failures: Logged and skipped (non-blocking)
- Invalid receipts: Logged with details, skipped
- Rate limiting: Exponential backoff (max 3 retries)

## Triggers

- `log llm usage`, `capture receipts`, `sync to lockn logger`