Skip to main content
Agents are autonomous AI workers that respond to events (triggers), reason about them using instructions you define, and take actions through tools. Each agent is scoped to an organization.

​
Agent lifecycle

StatusDescription
draftBeing configured. Not triggered automatically.
activeLive. Responds to configured triggers.
pausedTemporarily disabled. Configuration preserved.

​
Core concepts

Triggers define when an agent runs. An agent can have multiple triggers:
Trigger typeFires when
invoiceA new invoice is received or updated
email_receivedAn inbound email arrives
cronOn a recurring schedule
scheduledAt a specific scheduled time
slack_messageA Slack message matches criteria
Tools define what an agent can do. Each tool maps to a capability:
Tool typeAction
triageClassify and read entities
send_emailSend outbound emails
reply_to_emailReply to existing email threads
slackPost messages to Slack channels
trigger_webhookFire HTTP webhooks
create_taskCreate work tasks or request human approval
Runs are individual executions of an agent. Each run captures the full reasoning chain: goal definition, tool calls, results, and scratchpad notes. Runs can be triggered automatically (via triggers), manually, or in sandbox mode. Sandbox mode runs the real agent pipeline — same instructions, same LLM reasoning, same tool resolution — but intercepts side-effecting tools. Safe tools (reads, classification) execute normally. Guarded tools (email, Slack, tasks) return would_* payloads showing what the agent would have done without actually doing it.

​
Endpoints

List agents

All agents across the user’s organizations

Create agent

Create a new agent

Get agent

Full agent detail with triggers and tools

Update agent

Modify agent configuration

Delete agent

Permanently remove an agent

Trigger types

Available trigger types catalogue

Tool types

Available tool types catalogue

Manual run

Trigger an agent run

Sandbox run

Test an agent without side effects

Sandbox entity search

Search entities to use as sandbox input

List activities

Run history for an agent

Get activity

Detail for a specific run

Org activities

Run history across all agents in an org

Update scratchpad

Edit a run scratchpad