Developer guide¶
You're a developer if you're building agents that participate in a Yutha-governed swarm. That includes:
- Wrapping an existing LangGraph, CrewAI, OpenAI Agents, or Microsoft Agent Framework agent so it can join.
- Reading and writing envelopes through the SDK.
- Handling capabilities for tools that take consequential actions.
- Writing a new framework adapter against the spec.
If you're standing up and running the swarm itself, you're an operator — see the operator guide instead.
Start here¶
- Quickstart — the 15-minute joiner path. Bring an existing agent, get a passport, send and receive your first envelope, observe a capability check.
- Python SDK — the canonical client surface. Async by default; handles passport mint, bearer auth, envelope encoding, subscription multiplexing.
- Writing a new adapter — how to add support for a framework Yutha doesn't yet ship for.
Framework adapters¶
Four adapters ship today. Each has a full developer walkthrough and a paired worked example under Examples. Pick whichever framework matches what you already build agents in — the substrate is identical under all four.
- LangGraph — wrap a LangGraph node as a Yutha agent. The richest hands-on walkthrough; deterministic state-graph handler, no LLM dependency at the dispatch level. Paired example: customer support with refund cap.
- CrewAI — each
Agentin a Crew becomes a Yutha agent with its own identity. LLM-driven dispatch viaCrew.kickoff(). Paired example: AP & invoice processing. - OpenAI Agents — each
agents.Agentbecomes a Yutha agent. Handoff bridging viaRunHooksso every inter-agent transition produces a signed audit envelope; cap-gating via@capability_requiredonfunction_toolbodies. Paired example: research crew with citation enforcement. - Microsoft Agent Framework — each
agent_framework.Agentbecomes a Yutha agent. Cap-gating on async-native tool callables (no sync/async bridge); honest documentation of v1 scope vsWorkflowBuilder/RequestInfoExecutor/FunctionMiddlewarefollow-ons. Paired example: DevOps incident-response.