What it does
Requests a human has addressed to you: someone hit "Run now" on a task, or wrote @claude in a comment. A trigger is an explicit ask, so it outranks queue order — check this before next_task, and work what's here first. Each entry carries who asked, when, the task, and the instruction they typed (a mention's instruction is the comment itself). Open requests are listed whether or not they already reached you over the live stream: delivered_at records that a stream received it, not that anyone acted on it, so a request survives an agent that saw it and died. Claim one with claim_trigger before working it, so a second agent doesn't duplicate the work.
Parameters
| Name | Type | Required | Description |
|---|---|---|---|
limit | integer | Optional | Maximum requests to return (default 50).
|
Annotations
The hints the server sends with this tool. A host reads them to decide whether to ask you before a call.
- Read-only
readOnlyHint: true - Changes nothing.
- Destructive
destructiveHint: false - Only adds or reads; nothing already there is removed or overwritten.
- Idempotent
idempotentHint: — - Doesn't apply: the tool changes nothing.
- Open world
openWorldHint: false - Touches only your own Done account.
Which tokens can call it
A connection made with OAuth has full access. A token from Settings → Agent access → Extra MCP tokens can be narrowed, and a narrowed token never sees beyond its project or workspace.
- Full accessAllowed
- Read-only tokenAllowed
- Single-project tokenAllowed
- Single-workspace tokenAllowed
Try it
With Done connected, ask your assistant something like:
“Has anyone asked you for something directly?”
Connect Done to your assistant
One MCP endpoint per account, over OAuth or a token. Step-by-step setup for each client: