The green light, granted with proof.
A catalog of governed actions, for humans and agents.
The human-machine boundary, adjustable action by action.
A closed catalog
Agents call typed Actions by slug. There is no shell to reach.
Signed approvals
A human approves in the Channel the team already uses.
Execution stays in your network
A dispatcher runs each Job where you place it. Its credentials never leave it.
Every step is an Event
Who authorised what, where, with which rights: a query, not an investigation.
Job board
Group by plan127 Jobs
| Action | Platform / env | Member | State | Risk | Updated |
|---|---|---|---|---|---|
| k8s.rollout-restart | k8s-prod-cluster / staging | sre-agentAgent | QUEUED | Reversible | 2 minutes ago |
| queue.drain | notifications-svc / staging | sre-agentAgent | RUNNING | Reversible | 3 minutes ago |
| k8s.restart-pod | payments-api / production | sre-agentAgent | PROPOSED | Reversible | 6 minutes ago |
| db.adhoc-query | data-warehouse / dev | erinHuman | SUCCEEDED | Read-only | 12 minutes ago |
| disk.cleanup-docker | billing-batch / staging | sre-agentAgent | SUCCEEDED | Reversible | 20 minutes ago |
| cert.rotate | iam-gateway / staging | security-agentAgent | UNKNOWN | Reversible | 25 minutes ago |
| cache.warm | notifications-svc / qa | network-agentAgent | FAILED | Reversible | 1 hour ago |
| k8s.scale-replicas | payments-api / production | capacity-agentAgent | APPROVED | Reversible | 2 hours ago |
| db.vacuum-full | data-warehouse / production | erinHuman | REJECTED | Destructive | 3 hours ago |
| k8s.scale-replicas | search-api / qa | capacity-agentAgent | CANCELLED | Reversible | 5 hours ago |
Putting agents on infrastructure today means picking a bad option.
Direct access
The agent holds SSH, kubectl or cloud credentials. The only boundary is the prompt, which no auditor can bound.
Spectator chat
The agent suggests, a human copy-pastes. Safe, but nothing is recorded, reused or scaled.
Homegrown glue
Webhook, model, pipeline. Every team rebuilds the same plumbing, with no shared model of who may do what.
What is missing is the governed execution layer between agents and infrastructure.
The agent is a user.
Not a plugin, not a black box: a Principal under the same rules as its human colleagues. The boundary lives in the domain model and the network topology, where a prompt cannot move it.
billing-batch / staging
Proposed by
Approved by
Finished
Duration
An agent is a Principal
Its own identity, never the human's who set it up. A scope of Platforms and Environments, a budget, an autonomy level.
Read-only · 4
Automatic approvalauto-approvableidempotent
Modified 20 minutes ago by svc-catalog
Automatic approvalauto-approvableidempotent
Modified 20 minutes ago by svc-catalog
A closed catalog of typed Actions
Each Action is a reviewed script with a contract: inputs, risk, reversal, approval. The catalog is the whole attack surface.
Approval required
Two-person- ActionAutomatic
approval.minimumofk8s.scale-replicas - RiskAutomatic
reversible, idempotent
- EnvironmentDecidingTwo-person
productionsets a floor of two people - TeamAutomatic
paymentssets no floor here - AutonomyHuman
sre-agentatoperate: a protected environment needs a human
Signed approval in your Channel
The approval request arrives in your messaging tool with the justification file. The decision is signed by the human identity.
How a Job moves
Propose.
A member or an agent proposes a Job from the catalog. An agent attaches its justification file: trigger, diagnosis, evidence, confidence.Propose a Job
The Job enters the board as PROPOSED and waits for its approval — it is never launched from here.
- 1Target
- 2Action
- 3Parameters
- 4Schedule
- 5Change reference
- 6Reason
- 7Review
Target
billing-batch
stagingEvery state is named.
One lifecycle for human and agent work, and named states off the happy path too.The path of a Job that succeeds
- PROPOSED
- APPROVED
- QUEUED
- CLAIMED
- RUNNING
- SUCCEEDED
Approve.
The effective requirement is the strictest of every policy layer: automatic, one human, two people, or deny.Justification file
Trigger
Diagnosis
payments-api restarts every 4 minutes: memory reaches its limit first. A restart clears it until the fix ships.Evidence
- OOMKilled ×6 in 30 min
- memory 498Mi / 512Mi
Confidence
The model's confidence in its diagnosis. Recorded for you to read — it never lowers the approval requirement.
Dispatch.
A dispatcher placed in your network claims the Job and runs it. It exposes no API.Placement8 platforms · all served
production
payments-api
- dispatcher-prod-7f2c9Active
- dispatcher-prod-b41d0Active
- dispatcher-prod-c58e1Draining
- dispatcher-prod-e93a7Active
Audit.
Every transition is an Event on an append-only bus, attributed to a Principal.k8s.scale-replicas approved by erin, signed
InfoJOB_APPROVED
Proposal refused: db.vacuum-full is denied by policy in production
WarningPROPOSAL_REFUSED
sre-agent raised to operate by erin
InfoAUTONOMY_LEVEL_CHANGED
db.adhoc-query revised to r_0d45a8d9
InfoACTION_REVISED
Proposal refused: no dispatcher can run cache.warm on search-api
WarningPROPOSAL_REFUSED
Agent capacity-agent provisioned by erin
InfoPRINCIPAL_CREATED
Autonomy is demonstrated, not declared.
A remediation moves from free-form diagnosis to a cataloged Action, then to proposals a human approves, and only then to automatic approval. octroi.ai measures each step.
Autonomy levels, set by a human
0
observeProposes read-only Actions. A human approves each one.
1
assistRead-only Actions are approved automatically. Anything that changes state waits for a human.
2
operateReversible Actions are approved automatically outside protected Environments.
3
operate+Reversible Actions are approved automatically everywhere, with a change reference.
- Destructive Actions always wait for a human, at every level.
- Only a human raises a level, and the change is an Event.
- A failed Job the agent had approved automatically lowers its level by one.
The basics
Is it useful without AI agents?
Yes. Members use the same catalog, the same approvals and the same audit trail. Agents come later, as Principals under the same rules, when you decide.
Security
Does the agent hold credentials to my infrastructure?
No. An agent holds a token that lets it propose Jobs within its scope and budget. The credentials that reach your systems belong to the dispatcher, which runs inside your network and exposes no API.
Where are secrets stored?
With your dispatcher, in your own infrastructure, never in octroi.ai. The dispatcher hands them to the Job at run time and masks them in logs and outputs.
What stops a manipulated agent from doing damage?
The prompt is not the boundary. An agent can only propose Actions from the catalog, within its scope, and the state machine decides what waits for a human. A manipulated agent can propose; it cannot approve its own Job or step outside the catalog.