Skip to content

Run & Chat

ag2 run#

Execute an agent from a Python file or YAML config with a single message.

ag2 run <agent_file> --message "Your prompt here"

Options#

Flag Description
--message, -m The message to send to the agent
--max-turns Maximum conversation turns (default: 10)
--verbose, -v Show detailed agent output
--json Output structured JSON result

Examples#

Run an agent with a message:

ag2 run my_agent.py --message "What is the weather in SF?"

Pipe input from stdin:

echo "Summarize this" | ag2 run my_agent.py

Get structured JSON output:

ag2 run my_agent.py --message "Hello" --json

YAML Config#

You can define agents in YAML instead of Python:

config.yaml
llm:
  model: gpt-4o-mini

agents:
  - name: assistant
    system_message: "You are a helpful research assistant."

ag2 run config.yaml --message "Research quantum computing"

ag2 chat#

Start an interactive terminal chat session with an agent or team.

ag2 chat <agent_file>

Options#

Flag Description
--model Create an ad-hoc agent with this model
--system System message for ad-hoc agent
--verbose, -v Show detailed output
--max-turns Maximum turns per exchange (default: 10)

Examples#

Chat with an agent from a file:

ag2 chat my_agent.py

Create a quick ad-hoc agent:

ag2 chat --model gpt-4o-mini --system "You are a Python expert"

Session Commands#

During a chat session, you can use these commands:

Command Description
/quit End the session
/cost Show token usage and cost
/history Show conversation history

Session Recording#

All chat sessions are automatically saved to ~/.ag2/sessions/ for later replay and debugging. Use ag2 replay to review past sessions.