How do Background Agents work?

The Background Agent runs autonomously in the background - investigating an issue and prepping context before a human picks it up. It pulls account context, looks in connected MCPs, searches the knowledge base, docs, codebase, logs, runs skills, and more — so by the time you open an issue, the legwork is already done.

Background agents can be setup from the Agents page. They kickoff based on a specified condition and their output appears in the Assist Agent panel on an issue.

You can pick up where the Background Agent left off — ask follow-up questions, refine its findings, or use them to draft a reply.

The Background Agent will execute a skill (we recommend our out-of-the-box investigate skill). The agent has access to:

• Knowledge base

• Docs (through Training Data)

• Past issues

• Any codebase integrations you have connected

• Account context from fields and notebooks

• Any connected MCPs you give it access to

Setting it up

Go to the Background Agents page and configure a rule for when a background agent should run.

You can create different Background Agents for different purposes - for example you might have a different investigation run on billing issues.