ScoutML

ScoutML helps you track what your agent does during a run.

You add a decorator to a function like @track.code or @track.search, run your agent, and ScoutML saves those actions to the Sessions app.

Start Here

1. Install ScoutML
2. Set your API key
3. Run the quick start
4. Learn the decorators

5-Minute Setup

If you want the fastest possible path:

1. install ScoutML
2. set SCOUTML_API_KEY
3. run one script with track.init(...)
4. add @track.code and @track.search
5. call track.complete()
6. open:

https://scoutml.com/arg/research_sessions/

If you self-host ScoutML, use the same path on your own base URL.

What ScoutML Tracks

ScoutML focuses on action logging.

Common action types:

• code
• search
• tool
• note
• error

When you open a session in the app, you can see:

• the session task
• the agent name
• when the run started
• which actions happened
• how many times each action happened
• the payload for each action

First Example

from scoutml import track

track.init(task="Build a simple agent", agent_id="hello-agent")

@track.code
def write_code():
    return {"file": "main.py"}

@track.search
def search_web(query):
    return {"query": query, "results": 2}

write_code()
search_web("python decorators")
track.note("First run complete")
track.complete()

Need Help?

Go to Support.