Skip to main content

A2A Client

Connect to AgentVisor's A2A (Agent-to-Agent) transport from an external client — curl scripts and a Python client for Agent Card discovery, task creation, and streaming.

Difficulty: Intermediate

This is a client, not an agent — it demonstrates calling an already-running AgentVisor instance from outside the sandbox. Pair it with any agent that has A2A transport enabled.

What This Demonstrates​

  • Agent Card discovery: Fetching /.well-known/agent-card.json to find available skills
  • Task management: Creating, polling, and canceling tasks over JSON-RPC 2.0
  • Conversation continuity: Using contextId to continue a conversation across messages
  • Streaming: Receiving incremental updates via SSE

Prerequisites​

Start an agent with A2A transport enabled:

agentvisor template create langgraph/chatbot-agent
cd chatbot-agent
docker compose up -d
AGENTVISOR_API_A2A_ENABLED=true agentvisor serve .

Verify the Agent Card is being served:

curl -s http://localhost:8090/.well-known/agent-card.json | jq .

Setup​

In a separate directory:

agentvisor template create clients/a2a-client
cd a2a-client

Quick Start with curl​

# Run the full walkthrough: discovery, send message, poll status, list tasks,
# follow-up message, blocking mode, streaming request, cancel, error handling
./curl-examples.sh

# Or against a non-default host
BASE_URL=http://agentvisor:8090 ./curl-examples.sh

Python Client​

cd python-client
pip install -r requirements.txt

Discover agents:

python discover.py
python discover.py --json
python discover.py --url http://agentvisor:8090

Send messages and manage tasks:

# Send a single message
python invoke.py --skill chatbot --message "What is the capital of France?"

# Interactive chat session
python invoke.py --skill chatbot --interactive

# Continue a conversation
python invoke.py --skill chatbot --message "What about Germany?" --context-id <context-id>

# List tasks for a context
python invoke.py --list-tasks --context-id <context-id>

# Get or cancel a specific task
python invoke.py --get-task <task-id>
python invoke.py --cancel-task <task-id>

# Block until the task completes
python invoke.py --skill chatbot --message "Hello" --blocking

Streaming:

python stream.py --skill chatbot --message "Tell me a story"
python stream.py --skill chatbot --message "Continue" --context-id <context-id>

Files​

a2a-client/
├── curl-examples.sh            # Self-contained curl walkthrough of the A2A JSON-RPC API
└── python-client/
    ├── requirements.txt
    ├── discover.py              # Agent Card discovery
    ├── invoke.py                # Task creation and management
    └── stream.py                # Streaming task updates

See Also​