Skip to content
Search
Ctrl
K
Cancel
How it works
Compare
Docs
GitHub
npm
Select theme
Dark
Light
Auto
Menu
Start here
What is Realm, and when to use it?
Install and first run
How a run moves
Concepts
Workflows, steps and the run record
The four step kinds
Order and routing
Human gates and trust levels
Evidence, and how a run ends
Who drives a run
Guides
Write your first workflow
Call a service with an adapter
Write an agent step
Give an agent step tools
Add a human gate
Write a step handler
Run a workflow with realm agent
Connect an MCP client
Start runs from webhooks
Answer gates from Slack
Handle failure
Start runs safely
Test a workflow
Operate runs
Deploy a project
Let an assistant plan its own workflow
Upgrade Realm
Reference: workflow file
Workflow file: top-level fields
Workflow file: step fields
Input maps, template expressions and filters
Conditions: when, preconditions, abort_unless
Gates: trust and the gate block
Retry and timeouts
Agent-step controls
Services, tool servers, step templates, profiles and workflow context
Webhook trigger
JSON Schema blocks
What the loader refuses and warns about
Reference: command line
realm workflow
realm run: commands that read
realm run: commands that act
realm agent
realm listen
realm mcp and realm serve
Reference: MCP
MCP tools
Run state and health findings
Gate answers over MCP
Reference: code and settings
Adapters
Handlers
Deployment manifest (realm.yaml)
Cache economics file
Project extensions
Testing package
Core library: stores and the store contract
Reference: lookups
Error codes
Run record and export bundle format
Environment variables and files on disk
Glossary
How it works
Compare
Docs
GitHub
npm
Select theme
Dark
Light
Auto
Operating runs
This page has been replaced. Its contents are now on these pages:
Operate runs
realm run
: commands that read
realm run
: commands that act