Quickstart

1. Install the SDK

pip install futureagi

Python 3.9 or later is required. Beyond your existing LLM provider SDK, no dependencies are needed.

2. Wrap your client

Replace your provider import with the Future AGI wrapper. All other code stays the same.

# Before
from openai import OpenAI
client = OpenAI(api_key="sk-...")

# After: swap the import, add your Future AGI key
from futureagi import FutureAGI
client = FutureAGI(
    openai_api_key="sk-...",
    futureagi_api_key="fagi-..."
)

Find your Future AGI API key in dashboard Settings. Or set FUTUREAGI_API_KEY as an environment variable and leave it out of the constructor.

3. Add a rubric to your call

Include a rubric parameter in any chat completion call. It tells Future AGI what qualifies as a good response.

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Summarize this contract in plain language."}
    ],
    rubric={
        "criteria": "Factual, uses plain language, does not include legal jargon, under 5 sentences",
        "threshold": 0.8
    }
)

Set the minimum passing score with threshold (0 to 1). Calls scoring below it appear as dashboard flags and may trigger alerts.

4. Open dashboard traces

After one instrumented call, visit the Future AGI dashboard at gofutureagi.org and select Traces. You will see:

• Each call's complete prompt and response
• The rubric score (0-100) plus passed or failed criteria
• Latency, tokens, and model metadata
• Timeline view of multi-hop agent traces

LangChain integration

For LangChain, use the Future AGI callback handler, not a client wrapper. Add it to any chain or agent.

from langchain_openai import ChatOpenAI
from futureagi.integrations.langchain import FutureAGICallbackHandler

handler = FutureAGICallbackHandler(api_key="fagi-...")
llm = ChatOpenAI(model="gpt-4o", callbacks=[handler])

# All chains using this LLM are automatically traced
result = llm.invoke("What is the capital of France?")

LlamaIndex integration

With LlamaIndex, Future AGI observer uses LlamaIndex's built-in instrumentation hooks.

from llama_index.core import Settings
from futureagi.integrations.llamaindex import FutureAGIInstrumentation

fi = FutureAGIInstrumentation(api_key="fagi-...")
fi.register()  # hooks into global LlamaIndex instrumentation

# All subsequent LlamaIndex queries are traced

REST API reference

SDK actions use the REST API. Base URL: https://api.gofutureagi.org/v1. Send your key in the Authorization: Bearer header.

POST /traces

Submit a trace without the SDK. This suits environments lacking the Python or Node SDK.

{
  "trace_id": "your-uuid",
  "model": "gpt-4o",
  "prompt": "...",
  "response": "...",
  "rubric": { "criteria": "...", "threshold": 0.8 },
  "latency_ms": 340
}

GET /traces

Lists authenticated-account traces with pagination. Supports project_id, start, end, and below_threshold query parameters.