Developers / Agents

Agents use the same API people's clients use.

Two curl commands give an agent a list it can write to and a link a human can open. Scopes decide the rest.

Quickstart

Two commands. A list and a task.

Agent quickstart
Shared contract

One model for people, agents, and teams.

01 Person writes

A person types a task in any client. The line keeps its structure.

Prepare brief #!high #+agent #@launch
Person Writes a task in their own words.

Priority, owner, and context travel with the text.

02 Agent acts

The agent reads the list through the same API, scoped to the permissions it was granted.

{ "scope": "tasks:write", "audit": true }
Agent Operates within explicit scopes.

Every action is auditable. Scopes cannot self-escalate.

03 Team sees

The result appears in every client subscribed to the list. The original text and the agent's contribution are both visible.

Brief attached · source preserved
Team Sees the result with full provenance.

Agent work and human work appear in the same list.

Scopes and documented limits

Roles, permissions, and rate limits.

Role Read lists Write tasks Manage members Delete list Rate limit
Owner Yes Yes Yes Yes Documented limits
Editor Yes Yes No No Documented limits
Agent Yes Scoped No No Documented limits
Viewer Yes No No No Documented limits

Anonymous list creation is limited to 30 requests per hour and 200 per day. Agent-scoped tokens inherit the limits of the role that issued them. See the API reference for current numbers.

Give your agent an inbox.

Create a list, hand it a capability, and let it work.