Skip to main content

Background Agents

A background agent is an autonomous AI agent that runs on its own, with its own login and its own set of tools. It's not something you chat with directly like your normal AI client — it runs unattended (for example, a bot that files tickets or posts updates) and calls tools through Willow using its own credentials.

This page is your list of the agents you own or that your admin has assigned to you. Open it from Background Agents in the sidebar.

What you'll see​

Each agent appears as a card showing:

  • its name, slug, and description
  • a status badge (active or disabled)
  • the platform it runs on, if any
  • badges counting its tools, skills, and rules
  • when it was created

If you don't see any agents, either none have been assigned to you yet or you haven't created one. If you need an agent, contact your administrator.

Search and favourites​

  • Use the search box to filter agents by name, slug, or description.
  • Select the star on any card to add it to your favourites. Favourites sort to the top.
  • Use the Favourites button to show only starred agents.

Your favourites are remembered in your browser.

Creating an agent​

If your administrator has enabled it, a Create Agent button appears in the top right.

  1. Select Create Agent.
  2. Enter a Name (the slug is generated for you) and an optional Description.
  3. Select Create Agent.

When the agent is created, Willow shows its secret once in a confirmation message. Copy and save it immediately — it won't be shown again. If you don't see a Create button, agent creation is turned off for end users; ask your admin to create one for you.

Opening an agent​

Select any card to open the agent's detail page. It opens on the Overview tab and is organized into tabs.

Overview​

Summary counts for tools, skills, triggers, and rules, plus Recent Activity — a live feed of the agent's tool calls and trigger events. Select the eye icon on any entry to see the full details of that call.

Tools​

The tools the agent is allowed to use. For each tool you can see its owning app, a usage count, and a safety badge from Willow's guard checks.

If a tool needs you to sign in to the underlying app, it shows a Not connected badge and a Connect button. Select Connect, complete the sign-in popup, then reload to update the status. The agent can't use that tool until it's connected.

To ask for more tools, select Request More (see Requesting capabilities).

Skills​

The skills assigned to the agent. Skills are instructions that shape how the agent works; they don't call APIs. If none are assigned, you can select Request Skills.

Triggers​

Triggers run the agent automatically when an external event fires — for example, when a new issue is opened in a connected app. This tab shows when the agent was last triggered and a history of trigger events. Select Request More to ask for a new trigger.

Settings​

Connection details for the agent:

  • MCP URL: point your AI host at this URL to connect the agent over Streamable HTTP. Authenticate with the header Authorization: Bearer access_key:secret.
  • Download ZIP: download a ready-to-run bundle with the agent's configuration.
  • Credentials (only if your admin enabled it): view the access key, and rotate the secret. A rotated secret is shown once — save it right away. Rotating invalidates the old secret, so update anywhere the agent is configured.

Agents that run on a managed platform (such as Claude Managed Agents or AWS AgentCore) show that platform's connection card here instead of the MCP URL.

Requesting capabilities​

You can't add tools, skills, or triggers to an agent yourself — your administrator approves those. To ask for more:

  1. On the Tools, Skills, or Triggers tab, select Request More.
  2. For tools, browse or search the available list and check the ones you want. For any type, you can also describe what you need in your own words.
  3. Select Submit Request.

Your request goes to your administrator for review. Once they approve it, the new capability appears on the agent automatically.

Connecting an agent to your AI host​

  1. Open the agent and go to the Settings tab.
  2. Copy the MCP URL, or select Download ZIP for a pre-built bundle.
  3. In your AI host, add an MCP server using that URL and authenticate with Authorization: Bearer access_key:secret.

If the URL isn't shown, contact your administrator.

On this page