
Provision, pause, and operate Hermes on Google Cloud — no VPS, no installer to watch.
Recommended: connect from chat (MCP)
Use your hosted Hermes agent from Claude, Cursor, or Codex without leaving the chat window.- Open seaotter.dev/connect (signed in).
- One-click Add to Cursor — or copy the Claude / Codex command.
- Paste the onboard prompt into chat.
Alternative: API / scripts (curl)
1. Sign up and create an API key
- Create an account at seaotter.dev.
- Open Settings → API keys (org admin required).
- Create a key and copy the secret once. It looks like
so_….
2. List agents
3. Create an agent
201 with status: "provisioning" almost immediately. Provision continues asynchronously.
4. Poll provisioning stages
wizard_reachable is ok and status becomes running. See Provisioning lifecycle for status meanings.
Next
- What is SEAOTTER? — category definition
- Self-host vs managed — when to keep the VPS
- Connect from chat (MCP) — Claude, Cursor, Codex
- Authentication — API keys vs Clerk JWTs
- API reference — interactive try-it against production
- Upstream Hermes docs — tenant runtime, not the control plane
How long until the agent is running?
How long until the agent is running?
Typical wall time is a few minutes (usually under five) while namespace, Helm, pod, DNS, and dashboard gates complete. Poll
config.provisioning_stages until wizard_reachable is ok. Details: Provisioning lifecycle.Do I SSH into the box?
Do I SSH into the box?
No. Pause, restart, restore, secrets, and credentials go through the dashboard or API. See What is SEAOTTER?.
Is this seaotter.ai?
Is this seaotter.ai?
No. These docs are seaotter.dev (managed Hermes hosting). seaotter.ai is a different OtterScore product.
