Connect an agent
Sign up in the browser first, then connect a coding agent from an authenticated browser approval.
Start from Connected agents
Visit the signup page, create your account, verify email, and wait for approval if your account is in the signup queue.
After approval, open Connected agents and choose Connect an agent. Copy the instruction into your coding agent. When it shows you a pairing code and fingerprint, return to the browser to review and approve them. Public agent-first account creation is not supported.
What the agent may do
Once connected, a scoped trusted client may assist with ordinary onboarding work after your account exists. Sensitive handoffs stay in the browser.
- Prepare onboarding inputs — The agent may help describe the company goal and prepare setup answers. Template and package choices remain visible in the browser before approval.
- Request browser handoff — Provider OAuth, raw secret entry, payment, and first launch remain explicit browser steps owned by the signed-in human.
- Explain setup state — The agent can read scoped onboarding and runtime state, summarize what is missing, and ask you to complete browser-owned steps.
- Continue after launch — Launch progress remains visible in the browser. When it is done, the dashboard and agent can both read the resulting company state.
Supported agents
The connect-agent skill is a standard markdown protocol. Any tool-using coding agent can follow it. We develop and test against:
- Claude Code
- Codex
- Hermes
- OpenClaw
- OpenCode/opencode
If the conversation fails partway
Reconnect the agent from the signed-in browser UI. The account and signup approval are human-owned, so a failed agent conversation should not create a second account or a hidden draft.
If the platform itself returns an error during provisioning, the operator-clients API surface (see the API reference) returns structured error responses with retry guidance.