We make AI coding spend impossible to lose track of.
Ottto is the operating layer for AI coding cost: reconcile every agent, account, machine, and cloud bill into one trusted number - then act on it. Local-first, read-only, open by default.
One number you can trust - across every tool.
AI coding spend leaks across surfaces nobody owns: Claude Code subscriptions, Codex and Cursor seats, Pi credits, API keys, and Bedrock and Vertex bills - on every laptop and in the cloud. Each tool shows a sliver, none show the whole. Finance gets a number that doesn't match what engineers see.
Ottto reconciles all of it into one labeled, trusted picture - what was metered, what was estimated, and where it came from. We sit in the observability path, not your request path, so you get the number without changing how anyone works.
- ✓Every agent, account & machine. Claude Code, Codex, Cursor, Pi and more - subscriptions, usage, credits and cloud bills, reconciled.
- ✓Provenance on every dollar. Metered evidence and calculated estimates stay labeled, never blended into a guess.
- ✓From "what did we spend" to "what should we change". See the spend, then take the reviewed savings worth taking.
Principles, not just product.
These are the rules the product is built on. They are the reason engineers trust the number and let us near their accounts in the first place.
Out of your request path
We are observability, never a proxy. Ottto reads the telemetry your tools already emit plus admin-API data. It never sits in front of your agents, never holds your tokens, never adds latency. Provider auth changes do not put your agent traffic through us.
Provenance over guesses
Every number is labeled. Metered evidence reads as exact; modeled figures read as estimated. We never blend the two into a confident-looking guess. You always know how the picture was built, how fresh it is, and how complete.
You stay in control
Ottto recommends and acts only on reviewed actions you approve - never silent automation. We surface where the money goes and what to change; you decide what to take. Keep your tools, your plans, and your freedom.
Open by default
The runtime you install on your machine is open source - local-first, inspectable, yours. The provider-change feed behind Provider Impact is public too. You can read exactly what runs and what it touches.
The runtime is open source. So is the feed behind it.
The daemon that runs on your Mac is open and inspectable. The AI Provider Watch feed - which tracks every provider change that moves cost, quotas, token accounting, defaults, and deprecations - is a public project anyone can use.
Local-first observability for AI coding agents - see what Claude Code, Codex & other agents cost and do. This is the open runtime you install.
An open event feed and CLI for AI-provider changes that affect cost, quotas, token accounting, defaults, and deprecations. The engine behind Provider Impact.
Questions, design-partner interest, or a bigger rollout?
We are working closely with early teams shaping how AI coding spend gets measured and controlled. Whether you are kicking the tires solo or rolling Ottto out across an org, reach us directly.
See your real numbers in minutes.
No sales call. No proxy. No slowdown. Install the app, connect your accounts, and watch every agent, machine and credit reconcile into one trusted picture.