onllm-dev/onWatch

Track AI API quotas across Synthetic, Z.ai, Anthropic (Claude Code), Codex, GitHub Copilot & Antigravity in real time. Lightweight background daemon (<50MB RAM), SQLite storage, Material Design 3 dashboard. Zero telemetry.

onWatch – AI‑API quota monitoring made easy

What it does – onWatch is a lightweight, cross‑platform daemon (single binary < 50 MiB) that polls the usage APIs of many popular LLM‑backed services – Anthropic, OpenAI Codex, GitHub Copilot, MiniMax, Gemini, Cursor, Grok, Kimi, and more – stores the results in a local SQLite database, and serves a Material‑Design web dashboard (and optional menubar / VS Code status‑bar widget) so developers can see historical trends, upcoming reset times, and get alerts before they hit a quota limit or run out of budget.

Why it matters – Most providers only show a current snapshot of usage. onWatch fills the gap by keeping a per‑cycle history, calculating burn‑rate forecasts, and unifying dozens of providers into one view. This helps developers avoid unexpected throttling, plan experiments, and keep cloud‑spending under control.

Key features

  • Multi‑provider support – Synthetic, Z.ai, Anthropic, Codex, GitHub Copilot, MiniMax, Gemini CLI (legacy), Cursor, Grok, Kimi Code, Antigravity, Moonshot, DeepSeek, OpenCode Go, Mistral, Ollama Cloud, Muse, Command Code, plus custom JSONL API integrations.
  • Historical data – SQLite stores per‑poll usage; dashboard shows time‑series charts (1 h‑30 d) and cycle‑overview tables.
  • Alerts & notifications – Email, push (PWA), webhook, and on‑screen warnings when thresholds are crossed; configurable per‑quota.
  • Dashboard – Material Design 3 UI with dark/light mode, live countdown timers, burn‑rate projections, and a Prometheus /metrics endpoint (beta).
  • Desktop integration – System‑tray / menubar widget on macOS, Linux, Windows; VS Code extension shows the tightest quota in the status bar.
  • Zero telemetry – All data stays on the machine; the binary is self‑contained, no external services required.
  • Easy installation – One‑line script, Homebrew tap, Windows PowerShell installer, Docker image, or manual build from source (Go 1.25+).

Typical workflow

  1. Install the binary (or Docker image) and run onwatch setup to add API keys in ~/.onwatch/.env.
  2. Start the daemon (onwatch or via systemd/launchd). It polls each configured provider in parallel and writes usage rows to SQLite.
  3. Open http://localhost:9211 – log in with the admin credentials from .env – and explore the dashboard.
  4. Optionally enable email/push/webhook alerts and install the VS Code extension for in‑IDE visibility.

Who should use it – Individual developers, small teams, or anyone who runs code‑generation, code‑review, or other LLM‑driven workflows across multiple AI services and needs a single place to monitor consumption and avoid surprise throttling.

Project health – The repo shows GitHub stars, CI badges (coverage, downloads), a recent version tag (v2.14.11), and is listed in the Awesome Go collection. It is actively maintained (beta status) and provides extensive documentation for each provider.

Quick‑start summary

# macOS / Linux one‑liner
curl -fsSL https://raw.githubusercontent.com/onllm-dev/onwatch/main/install.sh | bash
# Windows PowerShell
irm https://raw.githubusercontent.com/onllm-dev/onwatch/main/install.ps1 | iex
# Then start the daemon
onwatch
# Open the UI
open http://localhost:9211   # or visit in a browser

Links – Website: https://onwatch.onllm.dev • VS Code Marketplace: onllm-dev.onwatch • GitHub: https://github.com/onllm-dev/onwatch

Related

  • Project
    langwatch/langwatch

    LangWatch is an open‑core platform for LLM observability, testing, routing, and governance. It lets teams trace every LLM call, run simulation tests, manage prompts, enforce budgeted virtual keys, and get usage reports across coding assistants, LangChain‑style agents, and major model providers.

  • Project
    ThinkWatchProject/ThinkWatch

    ThinkWatch is a self‑hosted gateway that centralises LLM API calls and AI‑tool (MCP) integrations for organisations. It authenticates via OIDC, enforces per‑user limits, applies security‑guard filters, logs every request to ClickHouse and provides a web console for configuration, cost‑tracking and audit. Deployable with Docker‑Compose or Helm, it supports OpenAI, Anthropic, Gemini, Azure, Bedrock and many third‑party tools.

  • Project
    open-webui/open-webui

    Open WebUI is a self-hosted AI platform that provides a user-friendly interface for interacting with AI models, supporting local models via Ollama and any OpenAI-compatible API. It offers features like plugins, role-based access control, persistent memory, RAG, and more, and can run entirely offline.

  • Project
    Willxup/cpa-usage-keeper

    CPA Usage Keeper is a self‑hosted dashboard that records, stores, and visualises usage data from the CLIProxyAPI (CPA) – an AI model proxy. It persists request logs in SQLite, shows cost breakdowns, latency metrics, and quota health, and can be deployed via Docker‑Compose, Homebrew, or native binaries.

  • Project
    verifywise-ai/verifywise

    VerifyWise is an open‑source AI‑governance platform that lets companies inventory models, evaluate LLMs, map risks to standards (EU AI Act, ISO 42001, NIST AI RMF, etc.), and generate audit‑ready reports. It runs a React front‑end, Node.js back‑end, PostgreSQL/Redis storage, and a Python EvalServer for LLM testing. Deployable on‑prem or in a private cloud via Docker/Kubernetes, it includes AI‑detection, policy management, Slack/Jira integrations, and an AI‑advisor chat.