aacp-crewai 0.1.0

AACP coordination layer for CrewAI multi-agent workflows

aacp-crewai

AACP coordination layer for CrewAI multi-agent workflows.

CrewAI maps more naturally to AACP than any other framework because agent roles map directly to AACP domains and tasks map to AACP task types.

Install

pip install aacp-crewai

Quick start

from aacp_crewai.crew import AACPCrew

crew = AACPCrew(model="gpt-4.1-mini")
result = crew.run_workflow("payroll", period="2026-03")
print(result.summary())

Measured results

Payroll workflow comparison. gpt-4.1-mini. June 2026.

                        WITHOUT AACP    WITH AACP
Coordination LLM calls:        4            0
Coordination cost:          $0.0005      $0.0000
Total cost:                 $0.0005      $0.0003
Total saving:                             39%
Coordination deterministic:    NO          YES
Schema validated:              NO          YES

The natural fit

CrewAI concept     AACP concept
──────────────     ────────────
Agent role         DOM  (HR, FIN, IT, SALES, CS)
Agent goal         Workflow objective
Task description   AACP packet content
Crew kickoff       Orchestrator run

Without AACP vs With AACP

Without aacp-crewai (standard CrewAI):
  Orchestrator
    ↓ "Retrieve all active employee salary records for March 2026.
       Include employee ID, name, department, cost centre, base salary,
       any changes this month, and pension rate. Return as JSON."
  HR Agent  ← verbose, varies every run

With aacp-crewai:
  Orchestrator
    ↓ FETCH|HR|return:ORCHESTRATOR|p:1|aacp:1.1|res:emp_salary|period:2026-03
  HR Agent  ✓ validates. $0.00 encoding. Identical every run.

Comparison demo

python3 examples/comparison.py --mock

Workflows

# Payroll (5 hops)
crew.run_workflow("payroll", period="2026-03")

# IT provisioning / JML (6 hops)
crew.run_workflow("it_provisioning", username="j.smith", dept="Engineering")

# Sales qualification (5 hops)
crew.run_workflow("sales_qualification", lead_id="L-001")

Requirements

• Python 3.10+
• OPENAI_API_KEY
• pip install aacp-crewai

Links

• Protocol spec: https://aacp.dev
• Python SDK: https://github.com/MackayAndrew/aacp
• LangChain integration: https://github.com/MackayAndrew/aacp-langchain
• Community rules (241): https://github.com/MackayAndrew/aacp-community-rules
• IETF Draft: https://datatracker.ietf.org/doc/draft-mackay-aacp/

Licence

MIT

aacp-crewai

AACP coordination layer for CrewAI multi-agent workflows.

CrewAI maps more naturally to AACP than any other framework because agent roles map directly to AACP domains and tasks map to AACP task types.

Install

pip install aacp-crewai

Quick start

from aacp_crewai.crew import AACPCrew

crew = AACPCrew(model="gpt-4.1-mini")
result = crew.run_workflow("payroll", period="2026-03")
print(result.summary())

Measured results

Department day comparison. 5 workflows. 59 coordination hops. gpt-4.1-mini. June 2026.

                        WITHOUT AACP    WITH AACP
Coordination LLM calls:       59            0
Coordination cost:          $0.0008      $0.0000
Agent cost:                 $0.0018      $0.0018
Total cost:                 $0.0025      $0.0018
Total saving:                             30%
Coordination deterministic:    NO          YES
Schema validated:              NO          YES
Audit trail structured:        NO          YES

CrewAI's natural language task descriptions are more verbose than LangChain's by default, which is why the per-hop saving is higher. Both comparisons use the same 59-hop department day scope for a fair comparison.

The natural fit

CrewAI concept     AACP concept
──────────────     ────────────
Agent role         DOM  (HR, FIN, IT, SALES, CS)
Agent goal         Workflow objective
Task description   AACP packet content
Crew kickoff       Orchestrator run

Without AACP vs With AACP

Without aacp-crewai (standard CrewAI):
  Orchestrator
    ↓ "Retrieve all active employee salary records for March 2026.
       Include employee ID, name, department, cost centre, base salary,
       any changes this month, and pension rate. Return as JSON."
  HR Agent  ← verbose, varies every run

With aacp-crewai:
  Orchestrator
    ↓ FETCH|HR|return:ORCHESTRATOR|p:1|aacp:1.1|res:emp_salary|period:2026-03
  HR Agent  ✓ validates. $0.00 encoding. Identical every run.

Comparison demo

python3 examples/comparison.py --mock

Workflows

# Payroll (5 hops)
crew.run_workflow("payroll", period="2026-03")

# IT provisioning / JML (6 hops)
crew.run_workflow("it_provisioning", username="j.smith", dept="Engineering")

# Sales qualification (5 hops)
crew.run_workflow("sales_qualification", lead_id="L-001")

Requirements

• Python 3.10+
• OPENAI_API_KEY
• pip install aacp-crewai

Links

• Protocol spec: https://aacp.dev
• Python SDK: https://github.com/MackayAndrew/aacp
• LangChain integration: https://github.com/MackayAndrew/aacp-langchain
• Community rules (241): https://github.com/MackayAndrew/aacp-community-rules
• IETF Draft: https://datatracker.ietf.org/doc/draft-mackay-aacp/

Licence

MIT