Skip to main content

Agents

Agents are the specialists that do the work in Studio: each has a system prompt, versions, and a live configuration that wires it to connectors, a knowledge graph and an execution environment. A fresh Studio has no agents; create and activate one before you open Chat.

Studio offers two types: "a Specialist focuses on a single domain and can be used in chat and in Workflows; a Supervisor works only inside Workflows." Supervisors "cannot have a knowledge graph, data connectors, or execution environment."

First run​

  1. Agent Management > Environments > Create Environment, Image Source Prebuilt (AlphaAgent Python); wait for Active (Execution environments).
  2. Create Agent (Specialist), Use as written or Optimise Prompt, Create Agent (v1) (Create and optimise).
  3. Open it, Configure > Set Environment > Save & Activate (Configure and activate).
  4. Open Chat (Chat).

The list​

  1. Open Agent Management in the dock. The list (/agents) is headed Manage Agents.
  2. Filter with Search by agent ID or name…; the search runs when you stop typing, across every page.
  3. Columns: Agent ID, Agent Name (newest version in brackets, Credit Analyst (v2)), Agent Type, Status (green Active or grey Inactive), Description (cut at 100 characters) and Updated At. Click a row or View to open it; Previous and Next page.

Where to go next​

PageRead it for
Create and optimiseThe Create Agent form, Use as written or Optimise Prompt, the view page, new versions.
Configure and activateThe Agent Map: connectors, knowledge graph version, environment, Save & Activate, deactivate, delete, the Model section.
Agent StoreDownloading a curated agent as a starting point.

Notes​

  • Each version is an immutable snapshot; the active version is what runs. Saving a new version changes nothing until you activate it.
  • Only agents with a live configuration appear in Chat's agent picker and in workflow nodes. An agent referenced by an active workflow version cannot be re-wired; create a new version instead.
  • A standard deployment holds up to 256 agents per user; at the limit Create Agent is disabled with "You've reached the limit of 256 agents." A governed deployment lists everyone's agents and the server enforces the limit on save.
  • "Error loading agents": the reason is underneath; reload.