Hangar

AI workloads · Autopilot

Any AI agent workload, bounded and audited

I think AI agents are going to become real teammates for platform teams, and they deserve the same paved road as people: the same contracts, the same checks, the same audit trail, just with less authority by default. Autopilot is Hangar's sixth product. It runs any agent workload on the platform that already runs the container apps, and it all comes back to one idea.

Durable changes are git commits. Ephemeral runs are claims to Crossplane.

Committing every agent session to git would be slow, noisy and impossible to review. So a run may skip git only if all five of these hold; if any fails, it goes through a pull request like everything else.

  1. It has a hard deadline that needs nothing else to enforce it.
  2. It holds no durable state, and reaches the durable plane only as a pull request.
  3. One narrowly scoped identity creates it, in one reserved namespace.
  4. It can only narrow what git grants, never widen it.
  5. It can be rebuilt from git plus the task spec.

Where it stands

Chapter 01

Where it fits

Hangar already had a rule: no component holds a live write credential, and every change is a git commit. Autopilot keeps that rule for anything durable and adds exactly one new kind of thing beside the existing platform.

Architecture · 01 of 19 · Hangar todayHangar as it is, and the one new component beside itOpen full page ↗
Architecture · 02 of 19 · Two planesDurable changes are commits. Ephemeral runs are claims.Open full page ↗
Secure paved road · 03 of 19 · The write pathThe durable path: an agent's write is a git writeOpen full page ↗

Chapter 02

What an agent workload is

Any agent workload is one of six shapes (task, session, service, scheduled, event, team), built from a durable, reviewed definition and a disposable run. A run gets a sandbox, a budget and a deadline, and it always ends, whether or not anyone is watching. A team is a tree in which every child is narrower than its parent.

Matrix · 04 of 19 · Workload shapesAny agent workload is one of six shapes, built from two kinds of thingOpen full page ↗
Architecture · 05 of 19 · Definition to runFrom a reviewed definition to a disposable runOpen full page ↗
Architecture · 06 of 19 · Run anatomyWhat a run gets, and the only two ways outOpen full page ↗
State machine · 07 of 19 · Run lifecycleA run always ends, whether or not anyone is watchingOpen full page ↗
Tree · 08 of 19 · Agent teamsAgent teams: every child is narrower than its parentOpen full page ↗

Chapter 03

Where it runs

Backends are managed the Hangar way: three fleet classes behind one contract, and two gates out of every sandbox, one for tools and one for models. Dev clusters get the write path and the runs; upper clusters get neither.

Architecture · 09 of 19 · Backend fleetsBackend infrastructure: three fleet classes, one contractOpen full page ↗
Architecture · 10 of 19 · Model and tool planeTwo gates out of the sandbox: one for tools, one for modelsOpen full page ↗
Deployment · 18 of 19 · Fleet placementWhere it runs: dev gets the write path and the runs, upper gets noneOpen full page ↗

Chapter 04

Identity and authority

An agent proves who it is on every call and gets a scoped token that is never kept. What it may do depends on its tool tier, and its effective authority is the minimum of five inputs, which runtime can only lower. Its pull requests meet every gate a human's do, plus two about agents.

Sequence · 11 of 19 · Identity and credentialsOne call, end to end: identity in, scoped token out, nothing keptOpen full page ↗
Access matrix · 12 of 19 · Tool tiersWhat each kind of agent may do, and who really does itOpen full page ↗
Architecture · 13 of 19 · AuthorityEffective authority is the minimum of five inputs, and runtime can only lower itOpen full page ↗
Flowchart · 14 of 19 · Agent PR guardrailsAn agent PR goes through the same gates, plus two that are about agentsOpen full page ↗

Chapter 05

Knowing what it did

The flight recorder reuses the telemetry Hangar already has and adds one tamper-evident hash chain. Preflight scores every agent change against incidents that have already been solved. The existing AI triage moves behind Clearance, so it gets the same bounds as everything else.

Data flow · 15 of 19 · Flight recorderFlight recorder: reuse the telemetry, add one tamper-evident chainOpen full page ↗
Process · 16 of 19 · PreflightPreflight: score every agent change against incidents you have already solvedOpen full page ↗
Architecture · 17 of 19 · AI triageAI triage: move Holmes's GitHub access behind ClearanceOpen full page ↗

Chapter 06

Getting there

Four rollout phases, each closed by a live check before the next begins, inside a plan of six milestones measured by one number: how operable Airframe is by agents.

Gantt · 19 of 19 · RolloutRollout: four phases, each closed by a live check before the next beginsOpen full page ↗
Gantt · 02 of 08 · RoadmapTwenty-eight weeks, six milestones, one score to moveOpen full page ↗
Dependency graph · 03 of 08 · OrderWhat blocks what: the critical path in six boxesOpen full page ↗

Chapter 07

The demo: Skyport

Skyport, Hangar's reference airline system, gains six AI workloads, one per shape. The event-shaped one turns a delayed flight into one draft, a team, and a human decision.

Matrix · 07 of 08 · Skyport AISkyport gains six AI workloads, one per shapeOpen full page ↗
Sequence · 08 of 08 · Event to teamA delayed flight becomes one draft, and a team, and a human decisionOpen full page ↗

Further reading

The reference set of ten diagrams (agent substrate, tool gateway, autonomy ladder, eval harness and more) is in thediagram gallery. The full design, with every claim labelled Built, Draft or Proposal, is in docs/autopilot/design.md.