Skip to content
Phlow Logo

A2A Protocol Framework for AI Agents

Phlow is a pure Python framework for building A2A (Agent-to-Agent) Protocol compliant AI agents. It provides authentication, agent discovery, multi-agent communication, and Supabase integration with real AI capabilities via Gemini API.

Get Started{ .md-button .md-button--primary } View on GitHub{ .md-button }

What is Phlow?

Phlow is a complete A2A Protocol implementation that enables autonomous AI agents to discover and communicate with each other. It includes JWT authentication, agent card discovery, real AI integration via Gemini, and comprehensive testing infrastructure.

Simple Integration

import os
from fastapi import Depends, FastAPI
from phlow import AgentCard, PhlowConfig, PhlowContext, PhlowMiddleware
from phlow.integrations.fastapi import FastAPIPhlowAuth

app = FastAPI()

# Configure your A2A agent
config = PhlowConfig(
    agent_card=AgentCard(
        name="My Agent",
        description="AI assistant agent",
        service_url="https://my-agent.com",
        skills=["chat", "analysis"],
        metadata={"agent_id": "my-agent-id"}
    ),
    private_key=os.getenv("PRIVATE_KEY"),
    supabase_url=os.getenv("SUPABASE_URL"),
    supabase_anon_key=os.getenv("SUPABASE_ANON_KEY")
)

middleware = PhlowMiddleware(config)
auth = FastAPIPhlowAuth(middleware)
auth.setup_agent_card_route(app)  # adds /.well-known/agent.json
auth_required = auth.create_auth_dependency()

@app.post("/api/analyze")
async def analyze(ctx: PhlowContext = Depends(auth_required)):
    return {"message": f"Hello {ctx.agent.name}"}

Key Features

🔐 A2A Authentication - JWT token verification with RSA keys

🛡️ Role-Based Access Control - W3C Verifiable Credentials for fine-grained authorization

🤖 Agent Discovery - /.well-known/agent.json endpoint compliance

🤝 Multi-Agent Communication - Agent-to-Agent task delegation

🧠 AI-Ready - Framework designed for AI agent integration (examples include Gemini)

📊 Supabase Integration - Agent registry and audit logging

🧪 Testing Infrastructure - E2E tests with Docker auto-detection

Architecture

graph TB A[Your Agent] --> B[Phlow Framework] B --> C[A2A Protocol] B --> D[Supabase Backend] B --> E[Gemini AI] C --> F[JWT Authentication] C --> G[Agent Discovery] C --> H[Multi-Agent Tasks] C --> I[RBAC & Credentials]

Getting Started

  1. Quick Start - Build your first A2A agent
  2. Installation - Setup guide
  3. Configuration - Environment variables
  4. RBAC Guide - Role-based access control
  5. Example Agent - Working code

Development

# See all tasks
uv run task --list

# Run tests
uv run task test-unit          # Unit tests
uv run task test-e2e-multi     # Multi-agent E2E tests

# Code quality
uv run task lint format type-check