Settings

Account

Sign out

Connected services

Connect an MCP server — Linear, Notion, your own — and its tools become available to your agents and apps. You sign in at the provider; the platform keeps the connection signed in, and your agents never see the credentials.

Connect

Connect an agent (MCP)

Point your own coding agent at yeebeebudev — the one MCP server to build, publish, and call the apps in this account. It acts as you, in your tenant.

Server URL

Claude Code — run this, then /mcp inside Claude Code and sign in when it asks:

Claude.ai / Claude Desktop — Settings → Connectors → Add custom connector, paste the server URL, then Connect and sign in.

Any other MCP client — add the URL as a remote (streamable HTTP) server. If it can't sign in with OAuth, create a coding-agent token on API tokens and paste the config it gives you.

Connected agents show up under API tokens (as OAuth clients); revoke one there to cut it off.

Browser notifications

Get notified in this browser when something needs your attention — even when the tab is closed.

Moderation queue

Images other people reported. Removing one deletes it everywhere and resolves its reports.

Admin console

System state, agents, and API tokens for this account.

DashboardAgentsEvolutionAPI tokens

Usage & cost

Estimated spend this billing month (UTC) — model turns, SMS, and platform meters. Tokens are the source of truth; dollars are derived from the current price book.

Permissions

What an app update wants that you have not agreed to yet — a new read, an external service, a public endpoint, a public page, or a self-serve sign-up door. It stays off until you allow it here; the rest of the app keeps working. Installing an app agrees to what its screen showed.

Cost controls

Hard limits on what your agents can spend — enforced by the platform, on top of any operator ceiling. Turns over a limit are shed (dropped and logged), never billed.

Save limits

Spend caps count model inference + provider messaging (SMS/push) — the costs that scale with agent activity. Egress and storage are metered and shown in your usage, but not capped (abuse there is bounded by rate limits). When a cap trips you get a typed reason naming the knob and when it resets.

Evolution → full page

How your events are handled as the system rewires itself — model turns vs self-authored reflexes vs narrow agents, read from the trace store. Full mechanics: handler chains, reflex versions, provenance spans, per-flow counters, and the daily fold-rate trend.

Timeline

Every change is a point on one timeline. Restore is non-destructive — you can always restore forward again.

• Loading…