atlas-agents-workshop

# Atlas agents workshop Connect, communicate and extend your agents. We build Atlas, a map agent that finds places near you, describes them from checked sources, and draws the walk to each one. One step at a time: an agent, tools, MCP, MCP auth, A2A and skills. You leave with an agent that has tools, an MCP server with a login, a second agent over A2A, a skill, and a map in front of them. ## Start here ``` git clone https://github.com/Kiruthika162003/atlas-agents-workshop.git ``` ``` cd atlas-agents-workshop/demo1-build ``` Then follow the [attendee guide](guide/Atlas-Attendee-Guide.pdf). It assumes nothing: it covers installing Python, Git and the Google Cloud CLI, creating a project with billing, signing in, checking the setup, and every step of the build with screenshots. Already set up? The short version is in [demo1-build/README.md](demo1-build/README.md). ## What is in the repository ``` demo1-build/ learn/ seven small agents, one new idea each (adk web) mcp_server/ the map tools as an MCP server, with an optional lock description_agent/ the second agent, served over A2A skills/ the walking tour skill atlas/ the web app and the recommendation agent check.py tests your setup and says what to fix start.py runs the whole app, on Gemini or on a local Gemma guide/ the attendee guide (PDF, and the HTML it is built from) ``` ## What you need - Python 3.11 or newer - Git - A Google Cloud project with billing on, and the `gcloud` CLI signed in - Optional: [Ollama](https://ollama.com) and a graphics card, to run it on a local Gemma model with no cloud account ## Next: put it in production The second session takes this same app to Cloud Run: a container, a private deploy, autoscaling, a load test, and chats stored in Firestore. It has its own repository and guide: [atlas-on-cloud-run](https://github.com/Kiruthika162003/atlas-on-cloud-run). ## Rebuilding the guide ``` pip install playwright ``` ``` python guide/make_pdf.py ```