TL;DR
In this guide, you will learn how to build full-stack Agent-to-Agent(A2A) communication between AI agents from different AI agent ...
For further actions, you may consider blocking this person and/or reporting abuse
Hi
What's up?
How many people have actually setup this project? Maybe I can get some help?
Hey what's up.
What's issues are you facing while trying to setup the project?
I wouldn't mind helping.
Hi Bonnie. I am trying to follow this tutorial. I am on windows 11. I performed the initial steps in this article to setup copilotkit and the research agent. Unfortunately, the scripts that are called when running in dev caused errors where the agents inside the 'a2a-demo/agents' folder failed to run. After some prompting, the best advise I received was that:
"scripts": {
"dev": "concurrently --names \"UI,Orch,Research,Analysis\" --prefix-colors \"cyan,gray,green,blue\" \"npm run dev:ui\" \"npm run dev:orchestrator\" \"npm run dev:research\" \"npm run dev:analysis\"",
"dev:ui": "next dev --turbopack",
"dev:orchestrator": "cross-env PYTHONUTF8=1 \"%INIT_CWD%\\agents\\.venv\\Scripts\\python.exe\" \"%INIT_CWD%\\agents\\orchestrator.py\"",
"dev:research": "cross-env PYTHONUTF8=1 \"%INIT_CWD%\\agents\\.venv\\Scripts\\python.exe\" \"%INIT_CWD%\\agents\\research_agent.py\"",
"dev:analysis": "cross-env PYTHONUTF8=1 \"%INIT_CWD%\\agents\\.venv\\Scripts\\python.exe\" \"%INIT_CWD%\\agents\\analysis_agent.py\"",
"build": "next build",
"start": "next start",
"lint": "next lint"
}
This works and the ochestrator.py and research_agent.py now run, but the analysis_agent.py still has an error due to uses the fact that the analysis_agent.py script uses google.adk.runners.Runner(...), which calls _enforce_app_name_alignment() and derives the app name from the directory it considers the app root. Since I activated the venv inside the agents folder and the cd .. — the launch path ends up being .venv.
I built out a batch script to run the initialization commands as follows:
`
@echo off
ECHO Starting the A2A Demo development environment...
ECHO.
REM This script assumes it is run from the project root (a2a-demo).
REM Step 1: Activate the Python virtual environment.
ECHO Activating Python virtual environment from 'agents.venv'...
call agents.venv\Scripts\activate.bat
IF %ERRORLEVEL% NEQ 0 (
ECHO.
ECHO ERROR: Failed to activate the virtual environment.
ECHO Please ensure it has been created at 'agents.venv' by running the setup steps in the README.
GOTO :EOF
)
ECHO Virtual environment activated successfully.
ECHO.
REM Step 2: Run the main development command to start the UI and all agents.
ECHO Starting all services with 'npm run dev'...
npm run dev
`
I have had the same issue too.
Set the google-adk version to 1.15.1 in the requirements.txt file then delete the .venv file and install the dependecies afresh.
Let me know if it fixes the issue.
Basically, I think the article needs a section on the considerations one should make for their development environment. It would be more confusing to switch to the docs to do a hard search, rather the info be part of this article to ensure that readers can follow along without environment setup issues.
How easy is this to set up? How long do you think it'll take me, Bonnie ?
Hey, Andrew and Steven, I followed Bonnie's setup process, and it's solid- took me about 30 minutes.
If you just clone the repo and add your ENV's it'll take just a few minutes, install, run the server, and away you go, talking to multiple agents.
Would love to get your feedback.
Ive been trying to set this project up, unfortunately the instructions here have not helped me as I am getting errors and am making modifications to the project structure and npm scripts that need to be ran. The article does not spare some time to atleast talk to dev environments and considerations. I am on Windows 11 and the relative pathing seems to not work well with powershell or cmd. Is it worth using wsl instead and the how to are some of the additions this article needs. A lot of content but we should still look away from the article is inconvenient.
Less than 5 minutes, Andrew.
+1
That's awesome, Steven.
Well all I can say, nice work—clean walkthrough and the agent handshake feels solid. A few tiny papercuts I hit: // … comments inside JSX (switch to {/* … */}) and a couple of unbalanced s. I also collapsed the 4 states into a reducer and lazy-loaded the card components so streaming stays smooth. Adding an aria-live="polite" on the results pane makes the stream accessible, and min-h-screen avoids the iOS viewport trap.
If helpful I can PR the small patch (imports tidy, reducer, lazy components, skeletons). Great job—this is a nice base for multi-agent travel UX.
Thanks for the recommendations, Damjan.
Let me look into it.
Nice Article 🙌
Thanks, Muhammad.
Nice! I’m looking for a post to learn AI-Agents.
Hi 👋,
Hope this post was helpful.
Niceee
Thanks, Henry
Cool stuff can't wait to try it myself. Thanks Bonnie
I am happy to hear that.
Awesome I love this handshake and great tutorial Bonnie was super easy to follow. Just got it setup myself
I am happy to hear that, Eli.
CopilotKit has been making big moves lately .. seems like they on a roll.
Nice job Bonnie for keeping up with them
The team there have been doing a great job.
AI may write the code, but understanding the logic behind it is what makes us true developers. The real skill now is guiding AI without losing our analytical edge.
nice
Thanks, Neha
Bonnie, this is so exciting!
A2A + AG-UI is pretty huge!
I learned a lot working on this guide.
Wow I've been waiting for this. Gonna try it out thanks Bonnie
I am happy to hear that, James.
Feel free to let me know how you find it out once you try it.
Cool! Saved for later
Thanks, Johny.