Guides
For the people
who implement it.
Your PM has designed AI agents and handed you the spec. These guides cover everything from reading the JSON to deploying the agents.
Reading an Envelope spec
What's in an .envelope.json file — agents, tools, triggers, and reporting lines. How to validate it against the open schema.
Deploying from a spec
Take an exported spec and wire up the agents. Which fields map to which runtime concepts, and how to handle tool credentials.
MCP host setup
Use Envelope's MCP endpoint as a zero-config deployment option. Add the URL to any MCP-compatible host and your agents run.
Storing specs in Git
Treat .envelope.json like any other infrastructure-as-code. Version it, diff it, review it in PRs.
Schema validation
How to validate an Envelope spec in your CI pipeline or IDE. JSON Schema tooling, VS Code setup, and common validation errors.
Tool credentials & permissions
How Envelope handles tool credentials in the spec. What stays in the file, what lives in a vault, and how to wire them at deploy time.
How agents are structured
The five components of every agent — role, prompt, model, tools, and reporting line — and how hierarchy, execution order, and gates fit together.
Looking for the full schema reference or MCP endpoint details?