Some screenshots show mcp-s.com, which some existing organizations still use. New organizations use withwillow.ai. Copy the URLs shown in your own organization.
Background agent organization settings
Settings govern every agent in the organization: which runtimes are available, and what owners and end users may do for themselves. Open Build > Background Agents and select the gear icon in the header. The page lives at /manage/background-agents/settings.
Changes are staged until you select Save Changes at the top of the page. Connecting a platform is the exception: those dialogs apply on Connect.
Agent Types
Chooses which platforms are available when creating an agent. Enable a platform, supply its credentials where one is needed, or define a custom type. Only enabled platforms are selectable in the create dialog, so this card is the source of truth for what your organization can pick.

| Platform | Setup |
|---|---|
| Willow Agents | None. Runs on Willow's infrastructure and draws on your Willow tokens. |
| Custom | None. |
| Claude Managed Agents | Set key, an ANTHROPIC_API_KEY. |
| Claude Tag | Optional Set key, a Claude Enterprise Admin API key for managing Slack access from Willow. Each agent is connected in Claude itself — see Connect Claude Tag. |
| AWS AgentCore | Set credentials. See Connect AWS AgentCore for the execution role and IAM permissions. |
| Cursor Cloud Agents | Set key for the Cursor cloud agent platform. |
| Snowflake Cortex Agents | Connect with the Snowflake account and token. See Snowflake setup. |
| kagent | Connect, the kagent controller base URL (and token if required). See Connect kagent for the SandboxAgent wiring. |
Platforms that need credentials stay switched off until you supply them. See Platforms for what each runtime does.
Add a custom agent type
Add custom type registers a runtime of your own, described as "the endpoints Willow calls to run this agent on your own platform." The dialog captures:
| Field | Purpose |
|---|---|
| Name | The platform's display name, for example My Runtime. |
| Description | What runs the agent. |
| Delivery | Willow calls your platform uses configured endpoints; Your platform collects work from Willow uses pull delivery without endpoint fields. |
| Sync endpoint | Called with the full agent snapshot when the Sync button runs. |
| Message endpoint | Called with the prompt whenever a trigger or message fires. |
| Request body (under each endpoint) | Optional. Willow default sends Willow's own body; Custom opens a JSON template where {{variables}} such as {{message}} or {{agent.slug}} are filled in per request. See Custom request bodies. |
| Auth header | The header Willow sends the token on. Defaults to Authorization. |
| Bearer token | Optional. Encrypted at rest, and sent as Bearer <token> when the header is Authorization. |
With pull delivery, Sync body and Message body configure queued payloads instead of endpoint request bodies. The Bearer token must match the value configured on the runtime; it identifies the type when the runtime collects work.
Create adds the type to Agent Types. Agents on a type created this way carry the custom-managed platform value.
For registration steps, see Register a custom agent platform. The delivered payloads are in the harness delivery reference.
Custom request bodies
Every endpoint field has a Request body control under it, set to Willow default until you change it. Switch it to Custom when your platform expects a shape of its own, and the editor opens pre-filled with Willow's body rewritten as a template:
{
"event": "{{event}}",
"agent_slug": "{{agent_slug}}",
"session_key": "{{session_key}}",
"message": "{{message}}"
}
Text outside a {{variable}} is sent literally, including constants such as a model ID or source flag.
| Control | What it does |
|---|---|
| Insert chips | Select a variable to drop it at the cursor. The chips list every value Willow can supply for that endpoint. |
| Build with AI | Describe the body your endpoint expects in plain language ("a prompt string and a thread object with an id") and Willow writes the template, wiring the variables for you. Requires Bring Your Own AI. |
| Preview | Shows the exact JSON your endpoint receives, with sample values filled in. |
| Reset to default | Returns the template to Willow's default body. |
| Use JSONata | Swaps the template for a JSONata expression evaluated against Willow's body. For reshaping that a template cannot express, such as building arrays. |
A variable used on its own as a value ("caps": "{{agent.capabilities}}") keeps its original type, so objects stay objects. A variable inside a longer string ("id": "willow:{{session_key}}") is inserted as text. A variable with no value at request time is sent as null.
Tool Assignment
| Setting | Effect |
|---|---|
| Allow common tools | Owners can assign tools from the organization's shared MCP servers. |
| Allow custom tools | Owners can generate and assign custom tools to their agents. |
Skill Assignment
| Setting | Effect |
|---|---|
| Allow common skills | Owners can assign skills from the organization's shared skill catalog. |
| Allow custom skills | Owners can create and assign custom skills to their agents. |
Credentials
| Setting | Effect |
|---|---|
| Show credentials to owners | Agent owners can view access keys and create, recreate, and revoke tokens from the connect dashboard. Off by default. |
End User Permissions
| Setting | Effect |
|---|---|
| Allow end users to create agents | End users can create their own agents from the connect dashboard. Off by default. |
| Require approval for new agents | When creation is allowed, end users request agents for administrator approval instead. Nothing runs until the request is approved and the necessary connections and deployment are complete. |
Save Changes applies staged organization settings.
These settings govern what agent owners and end users can do in the connect dashboard. The end-user view is documented in the user guide.
Channels
The Channels configurations card holds organization-level messaging configuration. Set up or Manage on Slack opens its connection settings. App-configuration tokens allow Willow to create each agent's dedicated Slack app.
| Deployment | What you do |
|---|---|
| SaaS | Store Slack app-configuration tokens (optional, enables one-click app creation per agent). |
| Hybrid | Register your slack-background-agent gateway URLs if self-hosting, then store app-configuration tokens. |
| On-prem | Store app-configuration tokens. The service URL is set by your deployment (SLACK_BG_AGENT_URL). |
The actual Slack app is created and installed per agent under Settings > Slack App. See Set up Slack for background agents for the full setup guide, runtime flow, and troubleshooting.
The card also has a Microsoft Teams row. Its dialog stores your organization's Azure Bot registration: the Microsoft App ID, the directory (tenant) ID, and a client secret. Unlike Slack, Teams uses one bot for the whole organization, so there's no per-agent app to create. This is a preview: agents don't answer in Teams yet. See Set up Microsoft Teams for background agents.
GitHub Apps (coding agents)
The GitHub Apps (coding agents) card on the Git tab stores the GitHub Apps that agents use to clone and push repositories. Add one App per GitHub organization your agents work in. It appears only when Git Proxy is enabled for your organization.
Under Token permissions, Contents must be write if any agent should push or open pull requests. Leaving it at read blocks those actions for every agent even when a repository is allowlisted as write on the agent. The GitHub App itself needs matching permissions. See Git Access for Coding Agents.