Overview
When you runstably create, stably build, or stably fix, the CLI automatically opens a session — a real-time view of what the agent is doing, visible from the Stably web dashboard.
Sessions let you:
- Watch progress live — see the current phase, activity log, and files being created or fixed
- Send messages to the agent — provide guidance, corrections, or additional context from the web UI while the CLI runs
create, build, and fix run.
Why Sessions Matter
In many environments you cannot interact with the CLI directly — CI pipelines, background agents, scheduled jobs, and remote machines. Sessions bridge that gap by giving you a web-based window into the agent and a way to communicate with it, regardless of where the CLI is running.Session creation is best-effort and non-blocking. If the connection fails, the CLI continues normally — your commands are never interrupted by a session issue.
How It Works
Run a CLI command
When you run
stably create, stably build, or stably fix, a session is created automatically — no extra flags needed.Watch progress live
Open the session on app.stably.ai to see what the agent is doing in real time — current phase, activity log, files being created, and more.
Send messages to guide the agent
Type a message from the dashboard and the agent receives it immediately. Useful when running in CI or other non-interactive environments.
Sending Messages to the Agent
From the web dashboard, you can type a message into any active session. The agent receives it in real time and adjusts its behavior accordingly — just like giving instructions in a terminal. This is especially useful in non-interactive environments like CI pipelines, Docker containers, or background jobs — places where you have no terminal access but still want to guide the agent. For example, you could send: “Focus on the checkout page, skip the login tests” and the agent will pick it up immediately.Session Types
Create / Build Sessions
Tracked phases:initializing → generating → testing → complete
What you see:
- The prompt or PR context being used
- Current agent activity
- Files being created
- Recent activity log with timestamps
- Environment info (git branch, working directory, CLI version)
Fix Sessions
Tracked phases:initializing → triage → fixing → validation → complete
What you see:
- Total, fixed, and failed test counts
- Which test is currently being worked on
- Individual worker status (for parallel fixes)
- Activity log per worker
- Link to the autoheal report on completion
Viewing Sessions
Sessions are visible on app.stably.ai under your project. Each session shows:- Live status — current phase and elapsed time
- Activity feed — chronological log of agent actions
- Message input — send messages to the running agent
fix command completes, the CLI also prints a direct link to the report:
Requirements
- Authentication — The CLI must be authenticated via
stably loginorSTABLY_API_KEY - Network access — The CLI needs outbound access to
api.stably.ai(HTTPS and WSS)