Envelope
an open schema for defining AI agent teams
Last verified:
What is Envelope?
Envelope is an open JSON Schema (Apache 2.0 licensed) for defining AI agent teams, enabling portable multi-agent system definitions that work across different frameworks and runtimes. The schema defines agent teams with roles, supervisor/sub-agent hierarchies, human-in-the-loop gates, pipelines, schedules, and secrets/variables injected at deploy time. It includes access policies that declare exactly which hosts each agent can call, enforced at the network level by the runtime rather than in prompts.
Key features include automatic VS Code autocomplete and validation via SchemaStore registration (no installation needed for .envelope.json files), programmatic validation through the @openenvelope/schema npm package, and a portable definition format similar to how Dockerfiles describe containers without being tied to a specific host. The spec covers agent definitions including role, prompt, model, and access policy configurations.
Envelope is designed for developers and teams building multi-agent AI systems who need a shared, framework-agnostic way to define agent teams. It targets those deploying multi-agent patterns with roles, handoffs, and human checkpoints who want their definitions to travel across different frameworks rather than being locked to a single tool. The managed runtime is available at openenvelope.org, but the schema itself is open for anyone to implement.
The schema is registered in SchemaStore and published on npm, making it immediately usable in development workflows. It solves the problem of scattered implementations by providing a canonical format for team definitions that any compatible runtime can execute.
Envelope pricing
Pricing model: Freemium
The schema itself is completely free under Apache 2.0 license with no cost to use, validate, or implement. The managed runtime at openenvelope.org pricing details are not publicly specified on the website - users need to contact the team for managed runtime pricing. Programmatic validation via @openenvelope/schema npm package is free.
Envelope pros
- Open standard with Apache 2.0 license for unrestricted use
- Portable definition format works across any compatible runtime
- VS Code autocomplete and validation via SchemaStore registration
- No installation needed for .envelope.json files in VS Code
- Programmatic validation available via @openenvelope/schema npm package
- Access policies enforced at network level, not just in prompts
- Defines complete supervisor/sub-agent hierarchy structure
- Supports human-in-the-loop gates for checkpoint control
- Includes pipelines and schedules for workflow automation
- Secrets and variables injected securely at deploy time
- Canonical schema lives at schema.openenvelope.org
- Anyone can implement their own runtime using the schema
- Solves framework lock-in by traveling across tools
- Complete spec covers role, prompt, model, and access policy
- Managed runtime available at openenvelope.org
Envelope cons
- Very new schema (released May 2026) with limited adoption
- Requires compatible runtime to execute team definitions
- Learning curve for teams new to multi-agent architectures
- Documentation may be incomplete given recent release
- Limited real-world case studies and examples available
- Network-level enforcement requires runtime support
- No built-in visualization or monitoring tools mentioned
- Schema version currently at v1, may have breaking changes
Frequently asked questions about Envelope
What is Envelope?
Envelope is an open JSON Schema for defining AI agent teams. It provides a portable definition format that describes agent teams with roles, supervisor/sub-agent hierarchy, access policies, human-in-the-loop gates, pipelines, schedules, and secrets/variables. The goal is to define a team once and have any compatible runtime execute it, similar to how Dockerfiles describe containers.
What is the license for the schema?
The Envelope schema is licensed under Apache 2.0, which means it is completely open and anyone can use, modify, and implement it without restrictions. The managed runtime exists at openenvelope.org, but the schema itself is open for anyone to implement.
How do I get autocomplete in VS Code?
The schema is registered in SchemaStore. Simply drop a .envelope.json file in VS Code and you get autocomplete and validation automatically without installing anything. You can also add a $schema reference to your file for automatic editor validation.
Can I validate team definitions programmatically?
Yes, the schema is published on npm as @openenvelope/schema, which you can use to validate team definitions programmatically in your applications or build pipelines.
What does the spec cover?
The spec covers agent definitions (role, prompt, model, access policy), supervisor/sub-agent hierarchy, human-in-the-loop gates, pipelines, schedules, and secrets/variables that get injected at deploy time. Access policies let you declare exactly which hosts each agent can call.
How are access policies enforced?
Access policies let you declare exactly which hosts each agent can call, and the runtime enforces this at the network level, not in the prompt. This provides actual security enforcement rather than just prompt-based constraints.
Do I need to use the managed runtime?
No, the schema is Apache 2.0 licensed and anyone can implement their own runtime. There is a managed runtime available at openenvelope.org, but you can use the schema with any compatible runtime you build or choose.
What problem does Envelope solve?
Envelope solves the problem that multi-agent systems have no shared way to define teams that travels across frameworks. Every implementation was scattered and locked to whichever tool was picked first. Envelope provides a portable definition format so you define a team once and any compatible runtime can execute it.
Is the schema version stable?
The current version is v1 of the Envelope Open Schema Spec. As a new open standard, it is the first official version. The schema lives at schema.openenvelope.org and is registered in SchemaStore for stability.
Who should use Envelope?
Envelope is for developers and teams building multi-agent AI systems who need a shared, framework-agnostic way to define agent teams. It targets those deploying multi-agent patterns with roles, handoffs, and human checkpoints who want portability across different frameworks rather than being locked to a single tool.