Plan / 01
Load context. Bound the task.
North Star reads the project goal and saved history, then defines one task with allowed paths, checks, a budget, and clear stop conditions.
Saved plan: project context, allowed changes, expected result.
From possibility to proof
Give ambitious work a direction.
North Star is a private, self-hosted AI command center being built to keep project context on your own server, schedule bounded tasks, and carry every result through human review into durable evidence.
Public preview · the working command center stays private
Plan Run Review Remember
The proof loop
One useful move.
Four visible stages.
Each stage shows what the server does, what it records, and where your judgment remains in control.
Run / 02
Schedule or start the worker.
The local server queues the mission for now or later, launches the chosen AI agent in an isolated workspace, and records each action.
Saved run: AI worker, limits, actions, and results.
Review / 03
Test the result before trust.
Independent checks compare source, tests, deployed behavior, and the original goal. Actions that need judgment wait for your decision.
Review record: findings, approval, real target proof.
Remember / 04
Save proof for the next run.
North Star writes the outcome, rollback path, remaining risk, and next useful action back into durable project memory.
Continuity record: what changed, what remains, where to resume.
A self-hosted AI command system
The work is the center.
Not the chat window.
The goal is to host the operator AI on your own server, give it the project context it needs, and schedule useful work inside explicit limits. Work, decisions, and results remain connected instead of disappearing into a chat transcript.
Right now
Prove the next useful release.
The current move is bounded. Review remains explicit.- 01Scope acceptedBoundary and checks recordedproved
- 02Real surface reviewInspect the result where it will livenow
- 03Carry the evidenceOutcome, rollback, next movenext
Needs your decision
One decision is yours.
Public visibility changes only after deliberate review.Latest evidence
Verification passedReal surface inspectedjust now
Designed public example · no live project data
Safety is architecture
A useful system knows where it must stop.
North Star does not pretend every decision can be automated. It keeps limits, review, and recovery visible as part of the route itself.
- 01
Clear limits
Every task starts with a project, a goal, a budget, allowed changes, and a clear point where it must stop.
- 02
Checked before trust
Changes travel with tests and pause for your decision wherever human judgment still matters.
- 03
Safe to undo
Rollback steps and results stay close enough to use instead of disappearing into a chat transcript.
Observation 05Built through real use
Current status
Private command center.
Active development.
North Star is being shaped through real project work: mission planning, review gates, evidence trails, rollback records, and a responsive operator interface. This page is the public story of that work, not access to the working system.
-
Posture
Private by design
The working command center stays behind authenticated access.
Protected boundary -
Direction
One useful move
Each mission has a visible edge, outcome, and stop condition.
Bounded route -
Memory
Evidence carried forward
Proof, rollback, and the next move survive the session.
Durable record -
Public surface
Preview only
This story contains no live controls or private project data.
Public safe
The closing horizon