bca2p is a Python SDK for bio-inspired agent-to-agent communication. It is designed to layer typed signaling, causal feedback, scoped communication, and adaptive topology on top of modern agent frameworks such as LangGraph and LangChain, while also supporting a native runtime path.
The repository now includes the stable V1 platform path plus experimental research tracks described in prd.md and agent.md.
Implemented layers:
- typed protocol models
- graph authoring and channels
- stable runtime and checkpointing
- causal learning and observability
- registry, transport, and A2A bridge
- LangGraph and LangChain integrations
- native runtime
- experimental MARL trainer
- experimental cell-signaling simulator
- experimental distributed substrate
src/bca2p/
core/ # Public protocol objects and schemas
graph/ # Graph builder and channels
runtime/ # Stable local runtime
learning/ # Causal inference and policy updates
transport/ # Local/remote transports and A2A bridge
registry/ # Agent and receptor discovery
observability/ # Traces, replay, diagnostics
integrations/ # LangGraph and LangChain adapters
native/ # Experimental native runtime
marl/ # Experimental communication training
sim/ # Experimental biology-faithful simulation
distributed/ # Experimental distributed substrate
- Primary target: Python
3.13.x - Secondary target: Python
3.14.x
The local workstation may use Python 3.14, but the package metadata and tooling are configured around a 3.13 baseline for SDK stability.
rufffor linting and formattingpyrightfor static typingpytestandpytest-asynciofor testspre-commitfor local quality gates
Create a virtual environment with Python 3.13, then install the package in editable mode with development dependencies:
python3.13 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"Run local checks:
ruff check .
ruff format --check .
pyright
pytest