Open Envelope proposes a portable schema for AI agent teams
Open Envelope is positioning an Apache-licensed schema for AI agent teams as a portable way to define roles, handoffs, schedules, access policies, and human approval gates.
What changed
Open Envelope is presenting an open JSON Schema for defining AI agent teams: roles, supervisor and sub-agent hierarchy, access policies, schedules, pipelines, secrets, and human-in-the-loop gates. The project says the schema is Apache 2.0 and intended to travel across compatible runtimes instead of staying locked to one orchestration platform. The Hacker News launch thread adds useful implementation detail: the author says the schema is available through schema.openenvelope.org, registered with SchemaStore, and published on npm as @openenvelope/schema.
Key takeaways
- Open Envelope is trying to define agent teams as portable configuration, closer to a reusable deployment artifact than a one-off prompt.
- The current scope includes agent roles, model selection, access policy, hierarchy, schedules, pipeline steps, secrets, and approval gates.
- The HN thread shows early pushback around declarative orchestration, trigger logic, access policy design, and whether gates should be modeled as first-class schema concepts.
- Source claims point to an Apache 2.0 schema, a managed runtime, SchemaStore registration, and npm validation support; each should be verified before production use.
- The cleanest LinkLoot use case is evaluation, not blind adoption: compare it against existing agent workflow formats before committing runtime work.
| Option | Best use | Limitation | Source |
|---|---|---|---|
| Open Envelope schema | Portable agent-team definitions with roles and gates | Young spec; runtime compatibility still needs proof | Open Envelope docs and HN launch |
| Claude Code workflows | Task-specific subagent orchestration inside Claude Code | Vendor-specific workflow model | HN discussion context |
| Custom internal schema | Exact fit for one team and stack | Harder to share across tools | Practical comparison |
Practical LinkLoot angle
Treat Open Envelope as a schema candidate for teams that already run repeatable agent workflows: customer support triage, research pipelines, migration assistants, or incident-response agents. The useful test is not whether the landing page sounds polished; it is whether the schema captures your real execution controls without moving policy into prompts.
A practical evaluation loop:
- Model one existing workflow with at least two sub-agents, one approval step, one secret, and one external host restriction.
- Validate the schema in a local editor and programmatically through the npm package if your stack uses Node.
- Compare the same workflow against your current tool, such as Claude Code workflows, a LangGraph-style graph, or an internal YAML format.
- Check whether access policies are enforced by runtime networking, not just described in natural language.
For more agent-tool evaluation patterns, use LinkLoot's guide to AI agent tools.
What to verify before you act
Verify the actual schema URL, npm package, license, and SchemaStore entry before basing production workflows on it. The HN discussion is useful for context, but it is not a release artifact. Also test whether access policy and human approval semantics survive a real runtime execution path; a schema that validates cleanly can still leave enforcement to code you have not audited.
Open Envelope is a project proposing a portable schema for defining AI agent teams, including roles, hierarchy, access policy, schedules, and approval gates.