Skip to main content
Beta: Agents is in beta and isn’t available in every workspace. To try it, contact the Pipedream team. Agent credentials can’t obtain tokens or access apps and tools yet.
Adding an agent doesn’t create or run one in Conduit. It gives an agent you already operate, such as a backend service or hosted assistant, a profile in your workspace, so Conduit knows which application is calling. The agent’s code, model and hosting stay where they are. The profile holds a client ID, credentials, and maintainers. Workspace admins can manage all agents; maintainers see the agents they’ve been added to.

​
Add an agent

  1. As a workspace admin, open Agents and select Create agent.
  2. Enter a name and optional description.
  3. Open the agent to add credentials and, optionally, maintainers.
Agents start disabled, with no permissions.

​
Credentials

An agent holds two credentials at a time, secrets and public keys together, so one stays in use while you deploy its replacement. A secret is shown once; a public key keeps its private key with you. Create a key pair with EC P-256 (recommended) or RSA 2048+, and keep private.pem in your secrets manager:
Paste public.pem or a JWK into Add public key. A pasted private key is rejected; treat it as exposed. The key ID shown under its label is the JWK’s kid or the key’s thumbprint. To rotate, add a replacement, deploy it, then revoke the old credential.

​
Maintainers

Admins can assign workspace members to maintain an agent: edit it, manage its credentials and disable it. Creating, deleting, enabling and assigning maintainers stay with admins. Removing a maintainer leaves the agent’s credentials in place; rotate any they may have kept.

​
Disable, delete and audit

Use an agent’s switch to disable it. Deleting an agent removes its credentials. The audit log records every change under Config → Agent.