aacp-pydantic-ai

aacp-pydantic-ai

AACP coordination layer for Pydantic AI multi-agent workflows.

Pydantic AI already types the result layer -- agents return validated Pydantic models. AACP adds typing to the instruction layer: what the orchestrator says to agents is now also typed, deterministic, and schema-validated. Both layers become fully deterministic.

Without AACP:  NL instruction → Agent → typed result
With AACP:     AACP packet    → Agent → typed result
               ↑ typed                  ↑ typed (already)

Install

pip install aacp-pydantic-ai

Quick start

from aacp_pydantic.orchestrator import AACPPydanticOrchestrator

orch = AACPPydanticOrchestrator(model="openai:gpt-4o-mini")
result = orch.run_workflow("payroll", period="2026-03")

Comparison demo

python3 examples/comparison.py --mock

Links

• Protocol: https://aacp.dev
• Python SDK: https://github.com/MackayAndrew/aacp
• IETF Draft: https://datatracker.ietf.org/doc/draft-mackay-aacp/

Licence

MIT