Home/Playground/Agent Workbench

Agent Workbench

A long-running agent flow with tool events, checkpoints, artifacts, and a handoff into review.

Why this pattern works

  • Tasks can change visual language when moving from execution to waiting or review.
  • Users see stage progression and artifacts at the same time, which avoids false certainty.
  • The surface keeps control chips visible so backgrounding and review never feel hidden.

Full event feed

Agent run queuedqueued

Provisioning a fresh execution environment.

Execution plan readyplanning

Five stages are visible before any write happens.

Planning executionplanning

Parsing the request and defining safe checkpoints.

Plan confirmedexecuting

Scope is locked and the workspace can be inspected.

Inspecting workspaceexecuting

Reading files and inferring the change boundary.

Tool startedexecuting

Search is scanning the repository for entry points.

Workspace mappedexecuting

The main routes and reusable primitives are identified.

Inspection completeexecuting

Safe edit boundaries are now visible.

Applying changesexecuting

The agent is editing and staging artifacts for review.

Artifact createdexecuting

A patch preview is available before verification.

Task moved to backgroundbackgrounded

The user can leave without losing progress.

Implementation checkpoint savedsynthesizing

The code changes are stable enough for verification.

Running checkssynthesizing

Build and smoke checks are underway.

Checks passedsynthesizing

The verification suite finished without blocking failures.

Ready for reviewready_for_review

Diffs, artifacts, and next steps are bundled for inspection.

Ready for Review02:02
Agent Workbench

2-10 minutes. This is no longer a loading state.

CancelBackgroundReview

Stage Rail

Plan executioncompleted

Lock scope, constraints, and success criteria.

Inspect workspacecompleted

Read files and infer the safe edit boundary.

Apply changescompleted

Mutate the project with checkpoints.

Run checkscompleted

Execute tests and gather proof.

Package outputpending

Collect diffs, screenshots, and review notes.

Artifacts

diff

Patch preview

A reviewable diff summary with ownership notes.

bundle

Review bundle

Diff, screenshots, logs, and verification output.

Evidence Feed

Ready for review02:02

Diffs, artifacts, and next steps are bundled for inspection.

Checks passed01:54

The verification suite finished without blocking failures.

Running checks01:34

Build and smoke checks are underway.

Implementation checkpoint saved01:32

The code changes are stable enough for verification.

Task moved to background01:10

The user can leave without losing progress.

Artifact created01:04

A patch preview is available before verification.