AI Usage

Check AI token usage, quotas, and costs across Anthropic and other providers using Anthropic OAuth API and OpenClaw session logs.

Security Scan

The skill's code and runtime instructions are coherent with its stated purpose (reading OpenClaw session logs and using the Claude/OAuth token to query Anthropic usage); it does access sensitive local credentials and invokes the local 'claude' CLI to refresh tokens, which is proportional to the task but worth reviewing before installation.

Purpose & Capability

The skill claims to report Anthropic quotas and OpenClaw-derived token/cost stats and its code reads ~/.openclaw session logs and the Claude credentials file and calls the Anthropic OAuth usage endpoint — these capabilities align with the stated purpose.

Instruction Scope

SKILL.md and the script instruct the agent to read OpenClaw session JSONL files and the Claude credentials file and to invoke the local 'claude' CLI to trigger an OAuth refresh. This is within the expected scope for retrieving usage/quota, but it does mean the skill reads a local credentials file and executes a local CLI.

Install Mechanism

No remote install/downloads or package installs are specified. The skill is instruction-only with an included Python script (no external network fetches for code), which minimizes installation risk.

Credentials

The script accesses the Claude credentials file and the OpenClaw sessions directory (both justified by the stated purpose). Minor metadata mismatch: registry lists no required env vars/primary credential, but SKILL.md documents optional overrides (OPENCLAW_SESSIONS_DIR and CLAUDE_CREDENTIALS_PATH).

Persistence & Privilege

The skill is not marked always:true, requests no elevated platform privileges, does not modify other skills, and only reads local logs/credentials and invokes a local CLI — no persistent/privileged behavior detected.

Assessment

This skill appears to do what it says: it reads OpenClaw session logs and your local Claude credentials to call Anthropic's OAuth usage endpoint. Before installing, review the included script yourself (it is present and readable) and confirm you trust the skill source. Note that it will read ~/.claude/.credentials.json (or the path set by CLAUDE_CREDENTIALS_PATH) and may invoke the local 'claude' CLI to refresh tokens; these actions are required to obtain accurate quota data but expose local credentials to the skill code. If you are uncomfortable, run the script manually in a restricted environment or inspect/modify it to remove any behavior you don't want (for example, printing or transmitting tokens). Also consider that the registry metadata omits the optional env var overrides documented in SKILL.md — expect to set those if your paths differ.

AI Usage Check

Quick Start

python3 scripts/usage_check.py           # Pretty report with gauges
python3 scripts/usage_check.py --json    # JSON output for scripting

Requirements

• Python 3.10+ (no pip dependencies — stdlib only)
• Claude Code installed and authenticated (claude CLI in PATH) — for Anthropic quota
• OpenClaw installed — for session log token/cost stats

What It Shows

Anthropic (real quota from OAuth API):

• Weekly utilization % with reset countdown
• 5-hour rolling window %
• Model-specific weekly (Sonnet, Opus) when available
• Extra usage spend vs monthly limit

Other Providers (from OpenClaw session logs):

• Token counts and call counts per model
• Auto-detects any provider OpenClaw routes through (Ollama, OpenAI, etc.)

OpenClaw Anthropic breakdown:

• Per-model token counts and equivalent API cost

How It Works

• Anthropic quota: GET https://api.anthropic.com/api/oauth/usage using Claude Code's OAuth token from ~/.claude/.credentials.json (requires user:profile scope)
• Token auto-refresh: If the token is expired, the script automatically refreshes it by invoking claude --print -p "ok" (Claude Code refreshes its own OAuth token on any invocation), then re-reads the updated credentials file. If Claude Code isn't installed, Anthropic quota is skipped gracefully.
• Session stats: Parses ~/.openclaw/agents/main/sessions/*.jsonl for per-provider/model token and cost data

Environment Variables

Variable	Default	Description
OPENCLAW_SESSIONS_DIR	~/.openclaw/agents/main/sessions	OpenClaw session log directory
CLAUDE_CREDENTIALS_PATH	~/.claude/.credentials.json	Claude Code credentials file

Tips

• Run via Haiku or a lightweight model for heartbeat/background checks — the script just needs to execute, no reasoning required
• Use --json for programmatic consumption (cron jobs, dashboards, alerts)
• Works with any Anthropic subscription tier (Pro, Max, Team, etc.)