Every dollar you spend on AI.
One cockpit. All local.
Meter and hard-cap every API call your apps and agents send through Omni's local proxy. See which model you're on in ChatGPT, Claude and Gemini. Keep Codex and Claude Code limits floating on top of your work. Usage is stored locally; provider calls and optional cloud features use the services you choose.
Preparing for macOS & Windows · downloads and purchases are not open yet
- Runs on your machineLocal proxy, database and dashboards
- Measured vs estimatedAPI metering is never blended with browser estimates
- macOS & WindowsBoth planned launch platforms
- $69 one-timeFounding License, first 20 customers
Your AI spend lives in six dashboards.
Your runaway agent doesn't check any of them.
Omni puts API keys, chat subscriptions and coding agents on one screen, and gives you the one thing the providers don't: a hard stop. Every number carries its source, so measured API usage, browser estimates and locally read agent limits are never mixed together.
A seatbelt for runaway agents
Point any SDK, agent or script at Omni's local proxy. Omni meters supported API traffic separately from browser estimates. With a hard cap enabled, a request is forwarded only when its worst-case cost can be bounded within the remaining budget.
- Monthly, per-provider, per-model and per-project limits
- Kill switch with a PIN, alerts before you hit the wall
- OpenAI, Anthropic, Google + 6 more providers through the proxy
# point your app or agent at Omni (with a cap on, also set max_tokens) client = OpenAI( base_url="http://localhost:8000/proxy/openai/v1" ) # Omni, a few hours later (illustrative) MONTH $ 49.97 / 50.00 CALL gpt-6.1-sol · 1.2M tok · $2.31 BLOCKED monthly limit reached · kill switch armed
What is your $200 plan actually worth?
The browser extension sees which model you're on in ChatGPT, Claude and Gemini, how much you use, and when a limit resets, then estimates its equivalent API value where a known model rate is available. These estimates are separate from measured API usage and are not provider invoices.
- Model and effort, when the page shows them
- Usage by 3h / 5h / 6h window, week and month
- Limit notices with reset countdowns
Codex and Claude Code limits, always in view
5-hour and weekly windows read straight from the agents' local session data, with countdowns to reset. Pin a manual timer for anything Omni can't read yet.
- Floating Ghost and HALO overlays stay on top while you code
- Per-agent usage percentage and next reset
A current model catalog
A curated, source-verified catalog of current models with real prices, context windows and release dates.
- Sort by newest, cost or context across every provider
- Unknown new models still show by their real name
Built so you never have to trust us
Runs on your machine
The proxy, database and dashboards all live on localhost. No Omni account and no cloud sync of your usage.
Keys stay in your vault
Provider keys are stored locally and only ever sent to the provider you're calling.
Unknown means unknown
Estimates are marked as estimates. If Omni can't measure something, it says so instead of inventing a number.
- Usage records for proxied API calls: provider, model, tokens, cost, latency, status
- Provider API keys, in a local vault
- Browser usage estimates, sent by the extension to Omni on localhost without prompt or response text
- Coding-agent limits, read from local Codex and Claude Code files
- Your own API calls, forwarded to the providers you choose, with your key
- License checks to Polar: license key, store identifier and a random install label, no usage data
- Optional cloud assistant and alert channels you configure
- No automatic analytics or crash reports to the developer
Real numbers and demo numbers never mix
Omni keeps three strictly separate modes, and every surface tells you which one you are looking at.
Your real telemetry
Measured API telemetry from the proxy, with separately labeled browser estimates. Demo data never enters Live. Live cockpit views need a license after the 14-day trial.
A labeled demo, free
Isolated simulated data, clearly marked DEMO, for seeing what Omni does before you connect anything. Free when the app is released.
A place to experiment
Isolated simulated data you can adjust to see how gauges, limits and alerts respond, without touching Live.
Try everything. Pay once if it earns its keep.
One honest offer while Omni earns its first reviews: a one-time Founding License for our first 20 customers. Purchases are not open yet.
The app’s Showcase mode uses simulated data, clearly labeled as demo. Free when the app is released; this website is not an interactive Showcase.
Release status- 14-day free trial first, no card, buy anytime
- 2 devices, any mix of Mac and Windows
- Keep every entitled version, permanently, even offline
- 12 months of updates from purchase, major versions included
- 14-day refund, no questions asked
Questions
When can I buy or download Omni?
Not yet. Customer downloads and purchases open once the macOS and Windows customer builds are ready. Launch status shows the current state, and support@omni-monitor.com answers questions in the meantime.
Does Omni see my prompts?
Proxied requests pass through Omni on your machine so it can count tokens and enforce limits. Your requests still go to the AI provider you choose. Optional cloud assistant and alert features contact their configured services. The browser extension processes page content locally to estimate usage and reads model and limit signals on chatgpt.com, claude.ai and gemini.google.com. It sends usage metadata to Omni on localhost.
Does the proxy slow my API calls down?
It runs on localhost and forwards the request directly to the provider; the overhead is a local hop.
What happens when the trial ends?
The Live cockpit asks for a license. Your apps keep working: Omni never blocks your proxied traffic because of a license. Showcase stays free. You can also buy during the trial.
Is the chat-app usage exact?
No, and Omni says so. ChatGPT, Claude and Gemini don't publish usage, so token counts from the browser are estimates and are labeled that way. API usage through the proxy is measured.
Mac or Windows?
Both macOS and Windows are planned launch platforms. Customer packages are not available yet. One license covers two supported devices in any mix once released.
Can I move my license?
Yes. Release a device from inside the app, then activate on the new one. Lost a device? Support will free the seat.
Does it need the internet?
API calls, cloud assistant features and configured online services need the internet. License activation and periodic re-checks also contact Polar. Once your license is verified, being offline or a license-server outage never locks you out.
Do I get updates?
Every release published within 12 months of your purchase, including major versions. After that you keep using what you have, permanently. There are no lifetime-update promises.
What if it isn't for me?
Ask for a refund within 14 days of purchase, no questions asked. The refund deactivates the key the next time Omni checks online; your API traffic, limits and local data are never touched. Refund terms.