Console tour

Watch your apps in the browser — deploys, logs, errors, data, secrets, settings, and the command palette.

The Console is read-mostly: it shows what your agent did to your apps and how they are used. Changes stay agent-driven. The Console edits data and settings, never code. Open it from the site header. Pick an app, or your account, in the sidebar. The command palette jumps anywhere.

One app

Overview

The app’s health at a glance: latest deploy, latest agent actions, open errors, and the request, visitor, error-rate, and p95 cards. When something looks wrong, this page tells you which of the views below to open.

Metrics

Requests over time with errors for comparison, daily or hourly buckets, top paths, and the same stat cards as Overview. Pick the range (?r=) to match the incident window.

Agent actions

Every tool call your agent made on this app: who ran it, what it did, and the outcome badge, newest first. When a deploy surprises you, this is the audit trail.

Deploys

Every deploy with its status badge, short SHA, and age. Expand one for the full SHA, findings, the live URL, and the error message for failures.

Open the Deploys view for your app.
Expand a deploy to read its findings and error message.
To undo one, pick an earlier successful deploy and confirm Roll back.

Rollback lives here: any earlier successful deploy has a Roll back button behind a confirm modal. Rollbacks never count as code changes. A “what changed” link jumps to the Code view’s changes tab for that SHA.

Code

Browse the app’s files at any SHA, and diff what a deploy changed (the changes tab, ?view=code&ct=changes). Read-only. Edits go through your agent.

Logs

The fleet’s stdout and stderr, newest first. Build output lives on the deploy, not here. For a failed deploy, expand it in Deploys or ask your agent for the log tail.

Errors

Crashes grouped by fingerprint with counts and last-seen times. These are the same groups your agent reads via app_errors. Triage (resolve, ignore, reopen) happens over MCP; the Console shows the state.

Secrets

API keys and tokens the app reads as Worker env vars (env.MY_KEY). Add, replace, or remove them here, or have your agent use set_secret/unset_secret. Names match ^[A-Z_][A-Z0-9_]{0,63}$, values hold at most 4 KiB, 50 per app. Values are never shown again after saving. Secrets apply on the next deploy. Setting one never touches the running fleet.

Settings

Slug, URL, status, your role, and last request. Then, owner-only:

Visibility — Unlisted (anyone with the link) or Public. Public is Pro (Who can see your app).
Custom domain — one hostname per app, Pro (Custom domains).
Collaborators — invite by email, up to 10 per app, Pro. Pending until they sign in (Collaborators).
Danger zone — rename (counts as a code change, redeploys) and delete (code, data and files go; the name stays reserved for 30 days).

Data

Owner-only browsers over the app’s live bindings:

D1 — the app’s database: tables and rows on the Tables tab (owners can insert, edit and delete rows), daily dumps on Backups (back up now, download a dump, restore one).
R2 — browse folders, preview text and images, download, upload, and delete. Reads use the same scoped credentials as the app; writes go through celld and are recorded in the agent action log.
Durable Objects — list instances (read-only). Selecting one shows what the app’s own ?read=1 handler returns. Ask your builder to add that read-only branch so you can preview state here.

Your account

Connections & keys — OAuth grants per client (revoke any time) and API key metadata: names and prefixes, never secrets.
Usage & plan — plan, usage against quota, reset dates (whoami shows the same over MCP). See Plans and Limits.
Account — email, terms, delete-account.
Invite codes — your OUTER-XXXX-XXXX codes during the closed beta: who redeemed them, which are spent. Admins generate codes in bulk on the Admin page instead and see the waitlist there.