Getting started

This page gets you from signing in to running an agent. Setting up the instance itself — connecting a repo, adding provider keys — is an admin task (see the admin introduction); once that’s done, everyone in the workspace can use the agents.

1. Sign in

Open the instance URL and sign in through your organization’s identity provider — Google, Microsoft Entra ID, or a generic OIDC provider (Okta, Auth0, Keycloak, …), whichever buttons appear on the login screen. A quickstart instance with no OAuth configured shows an email + password form instead. If it’s your first time, an admin may need to add you to a workspace.

2. Find your workspace

A workspace holds your team’s agents, runs, and connections. Use the workspace switcher at the top of the left sidebar to move between workspaces you belong to. The Agents list shows every agent in the workspace — the ones you built and the ones your teammates did.

3. Run an agent

Open any agent and click Run to execute it once. The run detail page shows a live, step-by-step timeline: the agent’s narration, the tools it called and how each turned out, plus token usage and cost. You can run agents other people on your team made — you don’t have to have built them.

4. Authorize your connections

If an agent talks to outside services (Slack, Gmail, Sheets, Attio, …), it acts as you when you run it — so authorize those accounts under Connections. Connections are per-user; each person authorizes their own. The sidebar prompts you to connect anything an agent needs that you haven’t yet. See Connections.

5. Automate and share it

You don’t have to run agents by hand:

• Schedule an agent or fire it from an event — Automations & triggers.
• Launch from Slack — Slack apps.
• Improve it from any run when the output isn’t right — Improvements.

These work on any agent in the workspace, including ones your teammates built.

Want to build your own?

Describe the agent you want in plain language and TAS opens a pull request via Tembo — see Authoring agents and Agent lifecycle. You can also commit a spec file by hand under agents/pydantic-agentspec/ in the connected repo.