Agent control plane
Orbit Harness
Put AI coding agents to work on real codebases — without handing over the keys.
Orbit Harness is a self-hosted control plane for AI coding agents from any vendor. Every agent works in an isolated workspace under hard budgets; the harness runs its own lint, tests and policy checks on the result; and nothing merges until a human approves a patch that is bound to its verification evidence.
What it does
Vendor-neutral by design
Agents plug in through standard task bundles and result schemas, so you can mix and switch providers.
Isolated workspaces
Every agent edit happens in its own git worktree; untrusted changes never touch your main tree.
Independent verification
The harness runs lint, tests and policy checks itself rather than trusting the agent's self-report.
Evidence-bound approval
A patch is hash-bound to its verification results and needs explicit human approval to apply.
Hard budgets
Token, cost, run, wall-clock and read limits are enforced, not suggested.
Secret hygiene and scoped reads
Credentials are stripped from agent environments and file access follows scope policy.
Clear lifecycle
Inspect, plan, implement, verify, review, apply — each mode with its own permissions.
Operable
Terminal UI, authenticated remote control plane and OpenTelemetry tracing.
Evidence
- 8,700+automated tests in the measured suite
- Lean 4formal specification workspace
- Mutationcoverage, SAST and mutation ratchets on the trusted core
Services around it
Production agentic systems
From prototype to production: we design and build agentic systems whose work is verified independently, bounded by budgets and approved by people before it lands.
Agent governance & evaluation
Already running agents? We assess what they can touch, build evaluation suites that measure what matters, and add the guardrails and audit trail you will be asked for.