Quickstart

1. Sign in

Sign in with Aitok OAuth. Your Aitok account is the compute wallet — model usage across AI Bee is metered against it.

2. Open the workspace

The workspace (/space) is where agents live. You get a personal space by default; create a team space when you need to collaborate.

3. Create an agent

Two ways to author an agent:

  1. Simple agent — fill in metadata and a set of prompts. This is the fastest path for prompt-only helpers.
  2. Smart agent — describe the goal in plain language and the AI builder generates the HTML/CSS/JavaScript plus a live preview. The agent runs on the Runtime with two database targets (shared server data and per-space client data) plus a file namespace.

4. Run it

Open the agent from the workbench. A smart agent renders in a sandboxed iframe and can call the Runtime:

// "server" is shared by every space (global rankings, cross-space data)
await runtime.db.query("server", "select * from public_scores order by score desc");
// "client" is private to the current space (local rankings, notes, settings)
await runtime.db.query("client", "select * from local_notes");
await runtime.file.read("client", "notes.txt");

Tables are created from migration/agent/*.sql and migration/space/*.sql. The two databases are separate and cannot be joined; for both a global and a per-space ranking, keep one table in server with a space_id column and filter by space_id for the local view.

5. Share it

Keep it in your space, or publish it to the market so others can install it. The per-space database and files are isolated per space; the agent's shared database is common to all installs.