Documents
README
README
Type
External
Status
Published
Created
Mar 17, 2026
Updated
Mar 17, 2026

with-a2a#

An example of using assistant-ui with the A2A (Agent-to-Agent) protocol.

Getting Started#

1. Start an A2A server#

You need an A2A-compatible agent server. For example, using the assistant-ui-a2a kitchen sink demo:

cd a2a-server
pip install -e .
python main.py

This starts an A2A server at http://localhost:9999 with 5 skills.

2. Configure the frontend#

cp .env.example .env
# Edit .env if your A2A server is not at localhost:9999

3. Run the frontend#

pnpm install
pnpm dev

Open http://localhost:3000.

Features Demonstrated#

FeatureSkill / Command
Streaming chatDefault chat
Artifacts (text, data, file)/artifacts <topic>
Multi-step input-required flow/multistep then provide topic
Error/failure handling/fail
Long-running + cancellation/slow
Agent card displayAutomatic on load
Task state tracking (8 states)All commands

Architecture#

Browser (Next.js) ──A2A v1.0 SSE──▶ A2A Server

No proxy backend needed. @assistant-ui/react-a2a handles the full A2A protocol directly:

  • Agent card discovery (/.well-known/agent-card.json)
  • Streaming via POST /message:stream (SSE)
  • Task lifecycle management
  • Artifact accumulation