Get started with Agent Stack

Step 1: install the Agent Stack

On MacOS and Linux you can install the Agent Stack cli by running the following script in the terminal:

sh -c "$(curl -LsSf https://raw.githubusercontent.com/i-am-bee/agentstack/HEAD/install.sh)"

for Windows follow full installation instructions

For this workshop, we will use OpenAI models, select the OpenAI LLM provider and configure your API key when requested during setup.

After a successful installation and setup, the command agentstack list should show the builtin example agents:

$ agentstack list
SHORT ID  NAME                    STATE     DESCRIPTION      INTERACTION  LOCATION                                               MISSING ENV  LAST ERROR
3b906a27  Chat                    missing   Agent with mem…  multi-turn   agents/chat:0.4.1-rc2                                  <none>       <none>
6b27a943  RAG                     missing   RAG agent that…  multi-turn   agents/rag:0.4.1-rc2                                   <none>       <none>
1af40f1e  Single-turn Form Agent  missing   Example demons…  multi-turn   agents/form:0.4.1-rc2                                  <none>       <none>

Step 2: clone the Agent starter template

The simplest way to create a new agent is to use the Agent Stack starter template.

After you create a new repository on GitHub, clone using git clone and start the agent using uv run server. See the README.md for repository structure and more details.

The agent should now show up in agentstack list as example_agent and also in agenstack ui. You can try sending a message using agentstack run <your-name>, you should get a polite greeting.

Now we are ready to modify the agent with our code, see the next part: 02-create-mcp-agent.