OnWatch is live · free · last verified 30 Oct 2026 · viability monitor
OnWatch
Open-source, local-first AI quota tracker for 16 coding-plan providers, with burn rate, reset cycles and per-session attribution in one dashboard.
If you pay for three or more coding plans at once, OnWatch is the quota tracker to install first: 16 providers in one localhost dashboard, automatic reset-cycle detection, burn-rate projection and per-session attribution in a Go binary idling under 50MB with zero telemetry. v2.14.1 adds a beta Prometheus /metrics endpoint for Grafana and Alertmanager, a menubar/tray companion and a VS Code extension. It is deliberately not a team platform — no SSO, no role-based access, no shared history across machines — and it stops at the 16 sources it supports. If your stack is a single Claude or Cursor subscription, the native counter is enough.
Verdict generatedVerified 3h ago · liveness 67/100 · cite: rightaichoice.com/tools/onwatch
- Developers running Claude Code, Codex, Copilot, Cursor and coding plans side by side
- Small teams sharing several AI accounts that want one headroom view without a cloud service
- Privacy-conscious users who want quota history and credentials to stay on their own machine
- Power users who want burn-rate projections and alerts before a quota runs dry
- Anyone paying for a single AI provider — the native dashboard is enough
- Teams that need SSO, role-based access or audit trails for quota management
- Buyers wanting a cloud-hosted, multi-user quota console rather than a local daemon
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip OnWatch if you only pay for one AI provider and just want to see a remaining-quota number, or if you need a cloud console with SSO, role-based access and shared team history rather than a daemon on your own machine.
Running it means doing your own ops: a local daemon to install, keep updated and back up on every machine, with no hosted version to fall back on
OnWatch is GPL-3.0 open source at no cost, so the comparison is not tier-versus-tier but build-versus-subscribe: you supply the machine and the maintenance, and in exchange you get burn-rate projection, reset-cycle detection and per-session attribution across all 16 tracked providers that a single provider's own dashboard does not show. For a solo developer on one subscription it is more setup than the native counter is worth; for anyone paying for three or more coding plans it is a fixed cost
In short
OnWatch — Open-source, local-first AI quota tracker for 16 coding-plan providers, with burn rate, reset cycles and per-session attribution in one dashboard. Best for Developers running Claude Code, Codex, Copilot, Cursor and coding plans side by side, Small teams sharing several AI accounts that want one headroom view without a cloud service, Privacy-conscious users who want quota history and credentials to stay on their own machine. Free to use.
What is OnWatch?
OnWatch — Open-source, local-first AI quota tracker for 16 coding-plan providers, with burn rate, reset cycles and per-session attribution in one dashboard. Best for Developers running Claude Code, Codex, Copilot, Cursor and coding plans side by side, Small teams sharing several AI accounts that want one headroom view without a cloud service, Privacy-conscious users who want quota history and credentials to stay on their own machine. Free to use.
OnWatch is a GPL-3.0 background daemon that runs on your own machine and tracks how fast you are burning through AI coding-plan quotas across 16 sources. Where a provider dashboard shows one number for right now, OnWatch keeps the history behind it: every poll lands in local SQLite, charted from 1 hour to 30 days, with automatically detected reset cycles closed out at peak and total consumed, burn rate projected forward to the next reset, and per-session attribution so you can see which agent run ate your weeklies. Coverage spans Anthropic's 5-hour, weekly all-model, per-model weekly, monthly and extra-usage buckets; Codex 5-hour, weekly and review requests with multi-account tracking; Antigravity's Claude, Gemini and GPT pools; Z.ai GLM Coding Plan credits; Synthetic; GitHub Copilot premium interactions; the MiniMax M2/M2.1/M2.5 shared pool; Cursor Individual, Team and Enterprise; Kimi Code; Grok; OpenRouter; OpenCode Go; Ollama Cloud; Moonshot balances; DeepSeek balances; and a legacy Gemini CLI source. A local JSONL drop-box lets you pipe bespoke API calls into the same charts. Credentials are mostly auto-detected from existing OAuth sessions, Cursor's SQLite store or the system keychain, and stored keys are encrypted at rest with AES-GCM. It ships as one Go binary under 50MB of RAM serving a Material Design 3 dashboard on localhost:9191, with beta SMTP alerts, Web Push over VAPID, a beta Prometheus /metrics endpoint, a menubar/tray companion and a VS Code extension. Current build is v2.14.1, installed via curl, Homebrew, PowerShell, Docker or Nix. This is a self-hosted developer tool, not a hosted team console.
Key features
- Quota tracking across 16 providers in one localhost dashboard
- Anthropic 5-hour, weekly all-model, per-model weekly, monthly and extra-usage buckets
- OAuth auto-refresh with token rotation to bypass usage-API rate limits
- Reads Claude Code's statusline when available to avoid the usage API entirely
- Codex 5-hour, weekly and review-request tracking with multi-account support (beta)
- Codex auto quota-starter that opens an unstarted 5-hour or weekly window
- Antigravity source switchable between desktop IDE probe, agy CLI, or both
- Z.ai GLM Coding Plan credit windows with per-tool breakdowns and tokens-per-call efficiency
- GitHub Copilot monthly premium-interaction and chat-completion reset tracking
- MiniMax M2/M2.1/M2.5 shared coding pool with multi-account support
- Cursor Individual, Team and Enterprise accounts with burn-rate forecasts and on-demand spend
- Kimi Code weekly window limits over OAuth with membership mapped to a readable plan name
- Grok Build and SuperGrok credit utilisation from the local CLI auth file
- Local JSONL drop-box to pipe bespoke API calls into the same charts
- Burn rate, reset-cycle detection and per-session attribution with 1h-30d trends
Behind the verdict
OnWatch solves a narrow problem properly. Provider dashboards answer "how much is left right now"; they do not tell you how fast you got there, whether this cycle is worse than the last, or which session burned it. OnWatch keeps every poll in local SQLite and layers four things on top: trends from 1 hour to 30 days, automatic reset-cycle detection that closes each cycle with peak and total consumed, burn rate extrapolated to the reset so you see the wall coming, and per-session attribution so each agent run becomes a comparable unit of work. There is also cross-quota correlation — which quotas spike together, and what 1% of your weekly costs you in 5-hour terms — which is the number that actually changes how you schedule a heavy run. The engineering choices are the reason it holds up. It is one Go binary under 50MB of RAM serving a Material Design 3 dashboard on localhost:9191, with zero telemetry and credentials in AES-GCM-encrypted storage. Most sources need no key at all: Anthropic, Codex, Kimi and Gemini CLI are read over OAuth auto-detection, Cursor from its desktop SQLite store or the system keychain, Grok from its local CLI auth file. Anthropic handling is notably careful — OAuth auto-refresh with token rotation to bypass usage-API rate limits, and it reads Claude Code's own statusline when available to skip the usage API entirely. Instrumentation is where the tool earns a permanent slot. Beta SMTP email alerts with per-quota threshold overrides on warning, critical and reset events; browser push over Web Push with VAPID; a beta Prometheus /metrics endpoint that drops into Grafana and Alertmanager; a menubar/tray companion; and a VS Code extension that puts your tightest quota in the status bar. The VS Code and tray pieces matter more than they sound — they move the check out of a browser tab you have to remember to open. The honest limits. Coverage is a fixed list of 16 sources plus the JSONL drop-box; if your stack runs on something outside it, you are writing normalised events by hand. Gemini CLI is kept working for existing setups but is no longer the recommended path — new installs should use the agy source on the Antigravity provider. Email and push alerts, multi-account Codex tracking and the macOS menubar app are all labelled beta. Because data lives in local SQLite there is no shared view across machines or teammates, no remote history, no mobile app, and no SSO, role-based access or audit trail. Credentials the tool stores are encrypted at rest, but many sources are read from tokens already sitting in other CLIs' local stores, so you inherit those tools' hygiene. Compared with watching each vendor tab in turn, OnWatch is the cross-provider view Claude Code's statusline and Copilot's monthly counter never give you — provided you accept a local daemon and its source list.
Models under the hood
as of 2026-10-25
Limitations
- OnWatch is a local-first, GPL-3.0 open-source daemon, not a hosted service.
- Coverage is a fixed list of 16 sources — Anthropic, Codex, Antigravity, Z.ai, Synthetic, GitHub Copilot, MiniMax, Cursor, Kimi Code, Grok, OpenRouter, OpenCode Go, Ollama Cloud, Moonshot, DeepSeek and a legacy Gemini CLI source — plus a local JSONL drop-box for your own calls; anything else has to go through the drop-box or is not read.
- Gemini CLI still works for existing setups but is no longer the recommended path; new installs should use the agy source on the Antigravity provider.
- Email and push alerts, multi-account Codex tracking and the macOS menubar app are all labelled beta, as is the Prometheus /metrics endpoint.
- Data lives in local SQLite, so there is no shared view across machines or teammates and no remote history.
- Credentials the tool stores are AES-256 encrypted at rest, but many sources are read from tokens already sitting in other CLIs' local stores, and the tool inherits that hygiene.
- The dashboard is served locally rather than being a cloud console, and there is no mobile app.
as of 2026-10-30
Researching OnWatch? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is OnWatch still active?
OnWatch is live · free · last verified 30 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is OnWatch? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What do users say about OnWatch?
What people actually say about OnWatch — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
Researched37 mentions across 5 sources (Hacker News, YouTube, Product Hunt, GitHub, Lemmy) · researched Aug 17, 2026.
Average across the 5 sources that answered — each source counts once, not each post.
- +Unified dashboard for 8 providers, ending dashboard-hopping
- +Very lightweight: under 50MB RAM, ideal for background daemon
- +Zero telemetry and local SQLite storage ensure privacy
- +Predictive burn-rate and reset countdowns help avoid overage
- +One-command install via curl, Homebrew, or PowerShell
- −Anthropic usage tracking unreliable in some cases (stalls)
- −No multi-account support for Codex, despite beta mention
- −Gemini CLI support missing, a popular provider for users
- −Alerts and menubar features still beta, potentially flaky
- −Limited provider coverage beyond eight—OpenCode gap
- • Setup time for multiple provider API tokens
- • Potential need to self-host for Docker users
How we verified this
Every listed fact about OnWatch— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 30 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified OnWatch 11 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed switching in, limitations, features, description, hidden costs, use cases, who should skip it, setup time, pricing fit, who it suits, our verdict, summary, switching out
- — changed pricing tiers, who it suits, description, our verdict, features, integrations, who should skip it, summary
- — changed SSL certificate
- — changed features, integrations, limitations, switching in, our verdict, switching out, who it suits, who should skip it, pricing tiers, pricing fit, setup time, description, use cases, summary, hidden costs
- — changed description, who should skip it, summary, our verdict, pricing tiers, integrations, features, who it suits
- — changed pricing page
Showing the 6 most recent of 11 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does OnWatch cost in 2026?
OnWatch: Free to use. Pricing last verified 30 Oct 2026.
Plans compared
Every published OnWatch tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free & Open Source | Free | — | Developers and small teams paying for several coding plans who want quota history, burn-rate projection and alerts on their own machineAdds: Starting tier and only tier: GPL-3.0, self-hosted, 16 provider sources, unlimited local SQLite history, zero telemetry |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where OnWatch's pricing actually pencils out — and where peers do it cheaper.
OnWatch is GPL-3.0 open source at no cost, so the comparison is not tier-versus-tier but build-versus-subscribe: you supply the machine and the maintenance, and in exchange you get burn-rate projection, reset-cycle detection and per-session attribution across all 16 tracked providers that a single provider's own dashboard does not show. For a solo developer on one subscription it is more setup than the native counter is worth; for anyone paying for three or more coding plans it is a fixed cost
Has OnWatch's pricing changed?
No plan price for OnWatch has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick OnWatch?
Pick OnWatch if you are Developers running Claude Code, Codex, Copilot, Cursor and coding plans side by side, Small teams sharing several AI accounts that want one headroom view without a cloud service, Privacy-conscious users who want quota history and credentials to stay on their own machine, Power users who want burn-rate projections and alerts before a quota runs dry.
Look elsewhere if you are Anyone paying for a single AI provider — the native dashboard is enough, Teams that need SSO, role-based access or audit trails for quota management, Buyers wanting a cloud-hosted, multi-user quota console rather than a local daemon, Users whose stack is outside the supported 16 sources, such as OpenAI direct or Vertex AI.
Skip it if Skip OnWatch if you only pay for one AI provider and just want to see a remaining-quota number, or if you need a cloud console with SSO, role-based access and shared team history rather than a daemon on your own machine.
GeneratedReal-world workflow fit
Concrete scenarios for the personas OnWatch actually fits — and what changes day-one when you adopt it.
You open localhost:9191 before launching a long agent run and check which of your Anthropic, Codex and Z.ai accounts has headroom left in its current window.
Outcome: You route the run to the account with the most remaining capacity instead of discovering mid-run that the 5-hour window was nearly exhausted.
The dashboard's burn-rate projection shows the current cycle running hot at a per-hour rate that will exhaust the weekly bucket before reset, and a beta SMTP alert fires on the warning threshold you set for that quota.
Outcome: You defer or batch non-urgent work into the next cycle, or switch providers, before the quota actually runs dry.
You point Prometheus at OnWatch's /metrics endpoint and add a panel and alert rule alongside your existing infrastructure dashboards.
Outcome: AI quota sits next to the rest of your operational metrics, and the VS Code extension keeps your tightest quota visible while you work.
Use cases
- Track remaining Anthropic Claude Code quota across 5-hour, weekly all-model and per-model weekly windows
- Compare headroom across 16 providers to decide which account to route a heavy agent run through
- Detect provider reset cycles automatically and plan batch jobs around the countdown
- Chart burn rate over 6 hours to catch a cycle running hot before it runs dry
- Attribute quota consumption per agent session to find which workload is eating your weeklies
- Monitor Synthetic hourly search limits with rate projections across all three quota types
- Feed bespoke API calls into the same charts through the local JSONL drop-box
- Run as a background daemon with one-command install on Mac, Linux, Windows or Docker
Setup time & first value
How long it actually takes to get something useful out of OnWatch — broken out by persona, not the marketing-page minute.
About 5 minutes for a solo developer: run the one-line curl installer (or Homebrew, PowerShell, Docker or Nix), let OAuth auto-detection pick up existing Claude Code, Codex or Cursor sessions, and open localhost:9191. Allow 15-30 minutes if you also want the beta Prometheus endpoint wired into Grafana and Alertmanager or the JSONL drop-box feeding custom calls.
Switching to or from OnWatch
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From checking each provider's dashboard by hand: install the daemon, let OAuth auto-detection populate the sources you already use, and let a few poll cycles build the first trend charts
- →From a legacy Gemini CLI setup: keep the existing source working, or switch new installs to the agy source on the Antigravity provider
- →From spreadsheets of quota resets: let automatic reset-cycle detection close each cycle with peak and total consumed instead of tracking dates manually
- ↗To a hosted team console: no export path is documented in the scraped content; your history lives in local SQLite
- ↗To native provider dashboards: stop the daemon and revert to one-tab-per-provider counters, losing burn-rate and reset history
Integrations
Tutorials & Learning
YouTube returned 6 videos for “OnWatch”, and we withheld 6: 6 could not be judged, because “OnWatch” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about OnWatch.
Official links
What are the best OnWatch alternatives?
The closest alternatives to OnWatch we list are Lilypad, Claw Lens and Langfuse. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Lilypad | Free | Free | — |
| Claw Lens | Free | Free | — |
| Langfuse | Freemium | Pricing not published | — |
Alternatives to OnWatch
View all
Tools that pair well with OnWatch
Common stack mates teams adopt alongside OnWatch, with the specific reason each pairing earns its keep.
Lilypad
Open-source OpenTelemetry tracing, versioning, and sessions for Python LLM apps—bring your own backend.
Claw Lens
Zero-config local dashboard that shows what your OpenClaw agents actually did, what they cost, and whether they leaked secrets
Langfuse
Open-source LLM observability: trace agents, manage prompts, and run evals on one MIT-licensed platform.
OnWatch vs its closest rival
OnWatch's closest head-to-head is with Spider Cloud — Spider Cloud is the better choice for developers building AI agents that need reliable, low-cost web data extraction, especially with its new Browser AI commands and 1,000+ scraper examples. OnWatch is uniquely valuable for heavy API users juggling multiple AI provider quotas, but it's a niche tool. For most AI workflows, Spider Cloud's Rust engine and broad integrations make it more versatile. Verdict generated
Head-to-head comparisons
Onwatch vs Spider Cloud
Spider Cloud is the better choice for developers building AI agents that need reliable, low-cost web data extraction, especially with its new Browser AI commands and 1,000+ scraper examples. OnWatch is uniquely valuable for heavy API users juggling multiple AI provider quotas, but it's a niche tool. For most AI workflows, Spider Cloud's Rust engine and broad integrations make it more versatile.
GeneratedOnwatch vs Temporal Ai
OnWatch and Temporal AI solve fundamentally different problems. OnWatch is a lightweight, free quota monitor for developers using multiple AI APIs, ideal for avoiding unexpected limits. Temporal AI is a powerful, paid-friendly durable execution platform for building reliable AI agents and workflows that survive failures. Choose OnWatch for quota visibility; choose Temporal for orchestration robustness.
GeneratedOnwatch vs Screenplayiq
Choose OnWatch if you're a developer juggling multiple AI API quotas and need lightweight, local tracking. Pick ScreenplayIQ if you're a screenwriter or producer who wants data-driven feedback on script marketability and box office potential. They solve completely different problems and are not direct competitors.
GeneratedFrequently asked questions
Used OnWatch? Help shape our editorial sentiment research.