Security
Checked on every change. On the record.
What every change passes before production, where your keys and data live, and the record that shows who decided what.
- Nothing reaches production without passing the line's checks: linting, tests, build, end-to-end runs and a security review on every single change. Machines check the machines; humans check the judgement.
- Bring your own model subscription or API keys, or run local models. They live with your orchestrator, never in the instance.
- Coding agents run on your laptop, a VPS or on-prem. The cloud instance never needs your source on our machines.
- Self-host: wherever you run it. Cloud: one private instance in Frankfurt, EU. AI calls go only to the models you route to; with local models, nothing leaves your premises.
- Export your whole instance at any time — it is yours either way.
Human in charge
Agents propose. Humans accept.
Every proposal, decision and test run is on the record, traceable to the requirement it served.
- Proposals
- Reviewed before merge
- Decisions
- Signed, dated, on record
- Evidence
- Bound to criteria, not to a chat log
The record — one requirement, one morning
| Time | Actor | Entry | Ref |
|---|---|---|---|
| 09:12 | agent-2 | Proposal: channel intake, no order counted twice | REQ-01 |
| 09:40 | Owner | Accepted, one change requested | REQ-01 |
| 10:05 | agent-2 | Workstep WS-01 done | WS-01 |
| 10:07 | test run #412 | 6 / 6 criteria pass | AC-01…06 |
| 10:09 | Owner | Released v1.0 | REL-01 |