Developer portal
Build on the control plane for AI work
Orcher sits between your people and your AI agents: every instruction is captured, interpreted, authorized, verified and recorded. Integrate your systems once — the governance comes with it.
Quick start
1
Create an organization
Sign in with your work email and provision your control plane. You get a sandbox with all seven components on, and a locked production environment.
2
Submit a directive
Use the playground or the SDK to capture an instruction. The AI interprets it, you confirm the interpretation, and approvers authorize it.
3
Run verified work
Once the Logic Scrubber returns a non-blocked verdict, tool calls execute under a signed authority token — and every step lands in the ledger.
The seven components
Directive Interface
Captures plain-language instructions and turns them into structured, risk-assessed interpretations.
Role Identity Fabric
Decides who may authorize what — per-role risk ceilings, two-person rules, signed expiring tokens.
Immutable Audit Ledger
A hash-chained record of every event, verifiable end to end and exportable as signed evidence.
Cost Governance
Budgets, model routing and metering for every AI call the platform makes.
Data Control Gateway
Allow / redact / block rules per data class, applied to everything entering or leaving.
Observability
Traces and spans across directives, model calls and tool calls.
Tool Gateway
Gated connections to email, CRM, ticketing, systems of record, data platforms and clouds.
Concepts
Directives, authority, environments, the ledger and data controls.
Read
API reference
Every callable operation: inputs, outputs, who may call it, what lands in the ledger.
Browse
Playground
Submit a sandbox directive and watch it move through the pipeline live.
Try it