How-to guides
Task-oriented, end-to-end walkthroughs of common jobs on the platform — follow one guide from start to finish instead of assembling steps from the reference pages.
The reference sections describe each part of the platform on its own — agents, skills, security, the gateway. How-to guides cut across them: each guide walks one complete task from start to finish, in order, and links into the reference pages where you want more depth on a single step.
Guides
Create an agent with the skill builder
Describe an agent in plain English, refine and test the drafted skill, configure everything, and watch the first build.
Register an MCP server in the Tool Registry
Admit an MCP server org-wide: pick the auth mode, review discovered operations, and open it up to skills.
Register an API from its OpenAPI spec
Turn an API's OpenAPI document into typed, governable tool operations agents can depend on.
Set up workspace service connections
Create named, tool-scoped workspace accounts for registered servers and bind skills to them.
Connect your personal account
Approve a user-delegated tool once — proactively, from a conversation, or over Slack — and let parked calls resume.
Govern a tool with compliance rules and PDP
Author value-level rules, tighten per agent, turn on fail-closed enforcement, and read the decision trail.
Approve deferred actions — console and Slack
Route deferrals to the right approvers, act on them from My Approvals or Slack, and understand timeouts.
Deploy a CI-built agent from your pipeline
Build the agent image in your own CI, deploy it with the CLI, and iterate with your pipeline in control.
Run an agent on a schedule
Make an agent autonomous, attach cron schedules, and verify each firing in Activity and Events.
Build a multi-agent workflow from a goal
Plan a pipeline from a plain-English goal, edit the canvas, run it, and monitor every step.
Lock down an agent's egress
Layer the org ceiling, workspace policy, and per-agent allowlist, then verify blocks in the event stream.
Your first agent
A short guided path — create an agent, watch it build, try it in a chat session, and read its activity trail.
Create an agent with the skill builder
An end-to-end walkthrough — describe an agent in plain English, refine the drafted skill in a conversation, review and test it, configure the agent, and watch the first build.