Reference · 12.01
Tool Registry: Build, Run & Prove
build_workflow, workflow.validate/dry_run/test/harden/bless, run.await/status/diagnose/replay, trace.inspect, and evaluate.
ToolsReference
Build
| Family | Tools |
|---|---|
| Build | build_workflow, plan_workflow, workflow.create/patch/validate/dry_run/test/harden/bless/restore_blueprint/scope/learn/patterns, agentis.workflow.graph.patch/replace, agentis.run.graph.evolve, workflow.delete (confirm-gated permanent removal + run history) |
workflow.patch is a whole-graph replacement contract: callers provide the complete graph. Use agentis.workflow.graph.patch for scoped stored-graph edits that preserve omitted fields, agentis.workflow.graph.replace for an explicit replacement, and agentis.run.graph.evolve for a live run.Run & prove
| Family | Tools |
|---|---|
| Run | workflow.run/status/deliver/cancel, run.await/status/inspect/query/diagnose/replay/regrade/cancel, workflow.loop_status |
| App proof | app.plan/doctor/compile — compile is the zero-side-effect admission proof for debug, production, and unattended execution. |
| Trace | trace.inspect, evaluate, ephemeral.run |
Prefer
run.await (event-driven, zero-token) over polling run.status in a loop. When only the acceptance contract changed, prefer run.regrade over another live execution — see The Execution Model.