Every run, in
its full context.
Launch and follow supervised agent processes with their workspaces, budgets, events, memory and evidence in view. Each surface connects back to the run that produced it.
Workspace migration review
Run execution map
Budget consumption
Recent events
- RUN ADMITTEDWorkspace + execution profile attached
- WORKER STARTEDResearcher · process p-042
- BUDGETToken allocation updated · 2,400 remaining
- HANDOFFEvidence review requested
Evidence report
One run. Six ways
to inspect it.
Move from process state to the context, limits and evidence behind a result.
A run groups a mission, its worker processes, coordination policy and execution receipts. The supervisor keeps each process attached to the workspace and budget that authorized it.
Explore surface ↗02 / ISOLATED WORKSPACESIsolated workspacesWorkspaces bound what an agent can inspect and change. Snapshots and isolated branches let candidate work proceed without silently overwriting the state that another path is using.
Explore surface ↗03 / BOUNDED EXECUTIONBudgets & limitsBudgets define how much time, work or model usage a run may consume. The supervisor can surface spend against its declared ceiling and make exhaustion an explicit execution outcome.
Explore surface ↗04 / OBSERVABLE EVENTSEvent historyA process view is useful only when it explains how the run reached its current state. Events make admission, handoffs, budget decisions and worker outcomes reviewable in order.
Explore surface ↗05 / SCOPED MEMORYScoped memoryRuntime memory can inform a mission with prior observations and decisions. Scope, source and retrieval context matter: an item should remain traceable to where it came from and why it was returned.
Explore surface ↗06 / REPORTS AND EVIDENCEEvidence reportsEvidence reports connect a run's claims to the artifacts, checks and provenance behind them. Reviewers can see what passed, what remains partial and which gaps still block promotion.
Explore surface ↗Operational visibility,
with clear boundaries.
This dashboard illustrates how run state, worker activity, budgets and evidence can be brought together. It uses sample values and does not connect to or launch a live GenOS process.
Read the runtime model →