Community
Grow at the edges, freeze the core
We build an open harness under the Metaorcha brand: one goal in, a verified multi-protocol run out, with an audit trail you can read. Not a chatbot framework, and not a closed agent marketplace.
The community adds protocol bridges, example agents, CanvasKit surfaces and docs. Handlers are pluggable, so a new protocol does not wait on a private roadmap.
Where to help
Protocol bridge
Subclass AgentHandler to connect a protocol such as LangGraph, OpenAPI or gRPC. A bridge can be written in under 50 lines of Python.
Example agent
Add a specialised agent with an agent manifest (emerge.yaml), its capability scopes and its DID (did:orcha:agent:*).
CanvasKit UI
Build declarative dashboard primitives, so a run renders structured widgets instead of chat bubbles.
Execution observer
Extend the ExecutionObserver seam and the validator to improve attestation and run evidence.
How to land a change
- Open a PR against the
communitybranch on solvent-labs-org/metaorcha. Maintainers review there and roll tomain. Prefer a labelled, scoped issue over an unscoped change. - Spec changes need an RFC issue before the PR. The agent manifest is frozen by default and versioned through schema_version.
- A bridge takes a dict payload in send_task() and returns a string. Prefix hard errors with Error:, and register the dispatch mapping and the emerge.yaml adapter.
Ground rules. The runtime runs in mock mode (PAYMENT_MODE=mock) with no proprietary cloud dependency. User agents use did:orcha:agent:*; platform tools use did:orcha:system:*. The CLI is orcha-sdk, the Python import is emerge, and trust-layer questions live on /kya.