​
Prerequisite

  • Python ≥ 3.10

​
Step 1: Install Agensight and Capture Your First Trace

Let’s start by installing the SDK and writing a small script to trace an OpenAI API call.

​
1. Create a virtual environment

python3 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

​
2. Install Agensight

pip install agensight

​
3. Write your first trace

Create a file my_llm_app.py:

from agensight import init, trace, span
import openai  # pip install openai

# Initialize Agensight
init(name="my-first-llm-project")

@trace("joke_generation_workflow")
def generate_a_joke():
    @span()
    def call_openai_for_joke():
        return openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=[{"role": "user", "content": "Tell me a fun fact about programming"}]
        )

    print("Calling LLM to get a fun fact...")
    response = call_openai_for_joke()
    fun_fact = response.choices[0].message.content
    print(f"Fun Fact: {fun_fact}")
    return fun_fact

if __name__ == "__main__":
    generate_a_joke()

Make sure your OPENAI_API_KEY is set before running the script:

export OPENAI_API_KEY=your-key
python my_llm_app.py

​
Step 2: View the Trace in Agensight

After running the script:

agensight view

This opens http://localhost:5001, where you can explore:

  • Token usage
  • Function spans
  • Full LLM call trace

​
Step 3: Set Up MCP Server with Cursor to Auto-Generate Config

To enable full tracing (agents, tools, prompts), you’ll need to generate an agensight.config.json for your project.

​
1. Clone and set up the MCP server

git clone git@github.com:PYPE-AI-MAIN/agensight_mcpserver.git
cd agensight_mcpserver

# Set up virtual environment
python -m venv mcp-env
source mcp-env/bin/activate  # On Windows: mcp-env\Scripts\activate

# Install dependencies
pip install -r requirements.txt

​
2. Connect MCP server to Cursor

In ~/.cursor/mcp.json, add:

{
  "mcpServers": {
    "agensight-server": {
      "command": "/path/to/agensight_mcpserver/mcp-env/bin/python",
      "args": ["/path/to/agensight_mcpserver/server.py"],
      "description": "Tool to generate agensight config"
    }
  }
}

Replace the paths with the actual paths to your Python binary and server.py file.

​
3. Ask Cursor to generate config

In Cursor, open your project and ask:

Please analyze this codebase using the generateAgensightConfig MCP tool

Cursor will run the MCP server and create a valid agensight.config.json in your project root.

​
What’s Next?

  • Add more @trace() and @span() decorators to trace your pipelines in detail.
  • Use the config file to enhance dashboard grouping for agents, tools, and prompts.
  • Learn about advanced features in the Core Concepts guide.