Self-hosted · local + cloud models · one governor

Your developers are running AI agents.
Who's governing them?

AOS is a self-hosted agentic AI service in miniature — the stack the frontier providers run internally, scaled to your hardware. It routes work across local and cloud models, bounds what every agent may spend, proves whether it did the work it claims, and rolls back what it didn't.

Runs on a laptop. No cloud dependency. Your keys, your models, your ledger. ·

Why AOS exists

Agent frameworks make agents run.
Nothing makes them accountable.

Q1

Did the agent actually do the work it claims it did?

Every run is verified against evidence AOS watched it produce — your build, your tests, and an independent fresh-context judge. "Looks done" is not a verdict.

Q2

What did that cost — and was it allowed to spend it?

Every run leases a hard budget before the first token, and settles its true cost to an append-only ledger. No lease, no run. No surprise invoice, ever.

Models and tools are interchangeable capabilities.
Governance is the durable asset.

What's in the box

A frontier-provider stack, on your hardware

Governed model routing

Local models (Ollama, LM Studio, vLLM) for routine work; Claude and OpenAI for the hard problems. Complexity, blast radius, and novelty decide the tier — and every routing decision is recorded and queryable.

Independent verification

Deterministic checks (your build, your tests), a sentinel watching every artifact, and a fresh-context model judge. A run passes on evidence, not on the agent's self-report.

$

Hard budgets & a real ledger

Runs lease money before they start and settle their true cost after. Budgets cascade to sub-agents and retries by construction — spending stops when the money does.

Containment & rollback

Each run executes in its own isolated git worktree. Rejected work is discarded entirely; verified work merges back. Your checkout is untouchable — always.

Earned trust

Autonomy is accrued, not assumed. Consecutive verified passes per task-type-and-model pair unlock longer leashes; failures revoke them. The trust ledger is the receipt.

Offline mode

Works on a plane: local-only routing that fails closed. In offline mode, nothing opens a socket to a cloud provider — guaranteed by the runtime, not by a config suggestion.

Teams, roles & approvals

Viewer, approver, operator, admin. Dangerous actions park on a human before they execute; every ledger entry records who launched the run. Break-glass admin included.

Full observability

A live console with an animated system map, provider health probes, per-run traces, Prometheus metrics, and one-screen diagnostics. Nothing phones home — telemetry is local-only, structurally.

How it works

Every run, governed end to end

Before a token is spent

AOS scores the goal, picks the cheapest model that can carry it, applies permission and risk policy, leases a hard budget, and checkpoints the workspace. A run that can't be verified or can't be afforded is refused up front — the cheapest failure is the one that never starts.

After the agent stops

The work is checked against your toolchain and judged by a model that never saw the agent's reasoning. Pass → merge, settle the true cost, add a trust point. Fail → roll it all back, settle what was actually burned, record why. Either way: an auditable entry, forever.

The call flow

One governor. Two outcomes: internal, or earned escalation.

Cloud Code, Code Easy, or any OpenAI-compatible tool makes the call — AOS decides where it runs. Routine work never leaves your hardware. Hard work is escalated, deliberately, with the reason on the record.

Cloud Code MCP · gateway Code Easy library · governed runner Any OpenAI client SDK → /v1/chat/completions AOS — the run governor 4 doors — library · CLI · gateway · MCP one governor behind all of them reserve — budget leased up front no lease, no run route — complexity · novelty · trust the cheapest model that can carry it floors — risk · blast radius · breaker dangerous work routes UP, never down verify — checks + fresh-context judge evidence, not the agent's self-report settle — true cost to the ledger who, what, cost, verdict — forever trust — earned, or revoked what ran well runs freer next time INTERNAL — your hardware Ollama · LM Studio · vLLM routine work · $0.00 on the ledger offline mode fails closed — nothing dials out, guaranteed ← most calls end here ESCALATED — the queen tier Claude · GPT · subscription CLIs complexity or risk earned this metered · budget-bounded · verified the reason is on the decision record ↑ only when it has to
The AOS console: account balance, a new-run form with detected verification commands, and live provider health — local Ollama and LM Studio up, cloud providers keyless, the claude subscription CLI available.
The console — launch a governed run, with live provider health: local models up, cloud keyless, the subscription CLI ready.
The AOS system map: connected hosts (Cloud Code over MCP and gateway, Code Easy over the library, an OpenAI SDK client), the governor's services (sentinel, checks, judge, breaker, approvals), and capability health per provider.
The live system map — every connected host (Cloud Code, Code Easy, raw SDK clients), every service, every capability. Driven by the record, not by guesswork.

Production integrity

Audited the way it will be attacked

AOS is built under a standing adversarial audit: independent reviewer passes over the money path, the wiring, and every API door — reporting only defects demonstrable from executable code. 618 automated tests gate every push, and every invariant below is verified against the code, not the documentation.

Reserve before the first tokenon every door — no lease, no provider call, ever.
Every reserve settlescomplete, abort, crash, or disconnect — partial usage settles at its metered cost, never $0, never forgotten.
One run, one ledger entryeven when a run outlives its lease and races the sweeper.
Offline stays offlinelocal-only mode fails closed before a socket ever opens.
The judge can't fail silenta crashed verifier is an abort, never a pass — and never a false rejection.
Sub-agents can't overdrawchild spend is bounded inside the parent's lease by a database constraint, not app logic.
Approvals are auditedevery human decision is recorded — and timeout means deny, exactly once.
Verified work is never droppedeven a merge conflict preserves it on its own branch, with your checkout untouched.

Integrations

Four front doors. One governor behind all of them.

aos run · CLI

The terminal front door: governed runs, budgets, trust, and a one-screen aos doctor. Five minutes from install to a verified run via aos setup.

aos serve · Console

The web console: launch and watch runs, approve dangerous actions, probe provider health, and see every host that's connected — live.

aos gateway · OpenAI-compatible API

Point any OpenAI-style client at /v1/chat/completions and every call becomes governed, metered inference. Existing tools keep working; the ledger starts.

aos mcp · MCP

Connect Claude Desktop — or any MCP client — and drive governed runs from chat, with approvals elicited in-conversation. Also a TypeScript library for embedding.

No API key? AOS can route through the Claude or Codex CLI on your own subscription — governed, metered, and budget-bounded like everything else.

Companion products

Better together

AOS governs any agent through its four doors — and two products already speak it natively.

Cloud Code CLI coding agent

A terminal-native coding agent with local and cloud models, permission gates, and MCP. Point it at AOS and every session becomes governed: connect over MCP to launch verified, budgeted runs from chat, or route its models through the AOS gateway for metered, ledgered inference.

Compatible with Cloud Code 1.3.0 and higher — MCP and gateway integration built in.

Code Easy autonomous build platform

Spec-driven autonomous development: capture a spec, let a multi-agent build loop implement and validate it. With AOS enabled, every work item runs under the governor — hard budgets, mid-run stops, real toolchain verification, and rollback when the work doesn't hold up.

Compatible with Code Easy 2.0.0 and higher — AOS governance selectable from the dashboard, API, or MCP.

Not using either? Anything that speaks OpenAI's API or MCP — or embeds a TypeScript library — plugs into the same governor.

Where it fits

Built for anyone who lets agents touch things that matter

The solo builder

Run coding agents overnight with a hard dollar ceiling and wake up to verified work — or a clean rollback and a receipt for exactly what was attempted. Local models keep the routine work free.

The engineering team

One governor for the whole team: per-user attribution, role-gated launches, human approval on dangerous actions, and a shared ledger that ends the "what did the AI spend this month?" mystery.

The home lab / air-gapped shop

A private, self-hosted agent service across your own machines — local GPU boxes for the work, cloud only when a problem earns it, and a fully-offline mode that fails closed.

The product embedding agents

Ship agentic features without shipping unbounded spend. Embed AOS as a library, or put the gateway in front of your existing runner — your product gets budgets, verification, and rollback for free.

Editions

Free where you're one person.
Paid where you're an organisation.

Solo

Free

  • The full single-user core
  • CLI, console, gateway & MCP
  • Local + cloud model routing
  • Budgets, verification, rollback
  • Offline mode

Team

Per seat

  • Everything in Solo
  • Users, roles & per-user attribution
  • Human approval workflows
  • Trust & decision analytics
  • Org-level model policy

Enterprise

Let's talk

  • Everything in Team
  • Multi-tenant deployment
  • Postgres & SSO
  • Custom policy & support

AOS never marks up your tokens. You bring your own keys and subscriptions; the product is the governance.

Stop trusting. Start verifying.

AOS is in private beta — v1.0.0-beta. Tell us what your agents run, and we'll get you a build and five minutes to your first governed run.

Request beta access

Self-hosted · your keys stay yours · nothing phones home