CrewAI
The closest conceptual match to the Envelope schema in the multi-agent landscape. Crew = team, agent = agent, kickoff = run. The translation layer is thinner than any other platform.
Overview
CrewAI is an open-source multi-agent framework with both a self-hosted Python library and a managed cloud offering (CrewAI Enterprise / AMP). Its crew model — a named group of agents with defined roles, tools, and a task flow — maps almost directly to an Envelope team definition.
The Envelope integration for CrewAI will use CrewAI's YAML-based crew definition format, converting the Envelope team definition into a valid CrewAI crew config at provision time. Research is complete; this is the next platform integration on the roadmap.
Concept mapping
| Envelope concept | CrewAI equivalent |
|---|---|
| Team | Crew |
| Agent | Agent |
| Run | Kickoff |
| Install | Crew provisioning |
| reportsToKey | Manager agent delegation |
| requiredSecrets | Environment variable secrets |
Why the fit is excellent
Most platform integrations require structural translation — mapping Envelope's flat agent list with reporting relationships into a platform's native format. CrewAI is different: the crew concept maps 1:1 to an Envelope team definition.
YAML-based definition — CrewAI's declarative YAML format closely mirrors the Envelope JSON schema — the conversion is mechanical, not interpretive.
Manager agent support — CrewAI's manager agent model maps directly to Envelope's reportsToKey hierarchy — manager agents delegate to specialists.
Task = Run — A CrewAI kickoff corresponds directly to an Envelope run invocation, including inputs, outputs, and completion status.
Open-source core — The self-hosted library means community-built adapters are feasible — developers can build CrewAI → Envelope converters independently.
Integration status
- [x] Research complete
- [x] Concept mapping confirmed
- [ ] Adapter implementation — in progress
- [ ] Conformance testing
- [ ] General availability