Documentation
Two kinds of reader.
Envelope is built for PMs who design AI agents, and for the engineers and IT leads who implement them. Pick your track.
For PMs & designers
Designing your AI agents
How to use Envelope to design, share, and export a structured AI agent design — no engineering required.
What is an agent spec?
How Envelope represents an AI agent design — roles, tools, reporting lines — as a structured JSON document.
Share a design
Generate a read-only link to share your team design with colleagues, stakeholders, or your IT team.
Export your spec
Download your agent spec as JSON. Open format, no lock-in, stores cleanly in Git.
Collaborate on a design
Invite teammates to your org so they can access, view, and edit shared designs.
For IT & engineers
Implementing a spec
How to take an Envelope export and deploy it — schema reference, MCP setup, and integration options.
Understanding the schema
The open JSON schema behind every Envelope export — fields, types, and validation rules.
MCP integration
Take an exported spec and wire it up to a Model Context Protocol host. Zero-config option available.
Deploying from a spec
How to take an .envelope.json export and implement it — which agents to run, which tools to connect.
Supported platforms
All orchestration platforms that run an Envelope AI agent system — capabilities, adapter coverage, and how to choose.
Looking for the MCP endpoint or full schema reference?