Agents in Leadtime are bot users that can receive work, act through the API, and leave a clear audit trail. Use them when work should be assigned to an automation or external agent instead of a person.
A bot is the identity in Leadtime. An agent connection defines how that bot works. A run or agent session is one concrete piece of work, such as a task assignment, a mention in a task comment, or a future automation.
Bot user: the Leadtime user record that appears in assignee pickers and audit history.
API tokens: credentials that let scripts, wrappers, and automations call the Leadtime Public API as that bot.
Agent sessions: signed webhook jobs created by Leadtime when it wants an external runtime to start work for a bot.
Provider connections: managed integrations such as Cursor Cloud or Claude Agent bots where Leadtime starts the provider session for you.
Self-hosted: you run the wrapper. Leadtime can issue bot API tokens and send signed agent-session webhooks. OpenClaw and Hermes are optional helpers for this type.
Cursor Cloud: Leadtime starts Cursor Cloud sessions for the bot and shows progress in the task history.
Claude Agent: Leadtime starts Anthropic-hosted Claude agent sessions and can attach skills, MCP servers, GitHub repositories, and Leadtime task tools.
Future native agents: will use the same bot/session model, but the runtime will be managed inside Leadtime.
When a task is assigned to a bot or the bot is mentioned in a task comment, Leadtime creates a new session for that event. The session appears in the task history as its own card. Reassigning the same bot later creates another session, so each request keeps its own history and result.
Self-hosted agents: use this when you build your own wrapper or connect OpenClaw or Hermes.
Bot API tokens: use this when a script or wrapper needs to call the Public API as a bot.
Cursor Cloud agents: use this for Cursor-hosted coding sessions.
Claude Agent bots: use this for Anthropic-hosted agents, MCP tools, skills, and GitHub repositories.
