Getting Started

1. Sign Up or Use Open-Source

Managed Service

Sign up and onboard at console.overmindlab.ai.

Self-hosted Open-Source Edition

Follow the readme for our open-source edition. This is fully local app that provides all key features of Overmind.

Regardless of your choice, the app will automatically create your project and API key and generate copy-paste integration snippet — no setup steps needed.

---

2. Install the SDK

Python

pip install overmind

JavaScript / TypeScript

# Pick the provider(s) you use
npm install @overmind-lab/trace-sdk openai            # OpenAI
npm install @overmind-lab/trace-sdk @anthropic-ai/sdk  # Anthropic
npm install @overmind-lab/trace-sdk @google/genai       # Google Gemini

---

3. Initialize Overmind SDK

Python

Call overmind.init() once at startup, before any LLM calls. Your existing LLM code stays exactly as-is.

Before:

import os
from openai import OpenAI

os.environ["OPENAI_API_KEY"] = 'sk-proj-'

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Explain quantum computing"}],
)

After:

import os
import overmind
from openai import OpenAI

os.environ["OVERMIND_API_KEY"] = 'ovr_'
os.environ["OPENAI_API_KEY"] = 'sk-proj-'

overmind.init(service_name="my-service", environment="production")

# existing code unchanged
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Explain quantum computing"}],
)

That’s it — add overmind.init() and set your API key. No proxy, no import changes, no key sharing.

Every LLM call is automatically traced (input, output, model, latency, token count, cost) and sent to Overmind. Your OpenAI and other provider keys stay in your environment.

Other LLM providers

Overmind supports OpenAI, Anthropic, Google Gemini, and Agno. Pass the providers you use:

import overmind

# Explicit providers
overmind.init(service_name="my-service", providers=["openai"])
overmind.init(service_name="my-service", providers=["anthropic"])
overmind.init(service_name="my-service", providers=["google"])
overmind.init(service_name="my-service", providers=["agno"])

# Or auto-detect all installed providers
overmind.init(service_name="my-service")

JavaScript / TypeScript

Initialize the Overmind client and call initTracing() before any LLM calls, passing the imported provider module:

OpenAI:

import { OpenAI } from "openai";
import { OvermindClient } from "@overmind-lab/trace-sdk";

const overmindClient = new OvermindClient({
  apiKey: process.env.OVERMIND_API_KEY!,
  appName: "my app",
});

overmindClient.initTracing({
  enableBatching: false,
  enabledProviders: { openai: OpenAI },
});

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

const response = await openai.chat.completions.create({
  model: "gpt-5-mini",
  messages: [{ role: "user", content: "Explain quantum computing" }],
});

await overmindClient.shutdown();

Anthropic:

import * as Anthropic from "@anthropic-ai/sdk";
import { OvermindClient } from "@overmind-lab/trace-sdk";

const overmindClient = new OvermindClient({
  apiKey: process.env.OVERMIND_API_KEY!,
  appName: "my app",
});

overmindClient.initTracing({
  enableBatching: false,
  enabledProviders: { anthropic: Anthropic },
});

const client = new Anthropic.default({ apiKey: process.env.ANTHROPIC_API_KEY });

const message = await client.messages.create({
  model: "claude-sonnet-4-20250514",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Explain quantum computing" }],
});

console.log(message.content[0].text);

await overmindClient.shutdown();

Google Gemini:

import * as GoogleGenAI from "@google/genai";
import { OvermindClient } from "@overmind-lab/trace-sdk";

const overmindClient = new OvermindClient({
  apiKey: process.env.OVERMIND_API_KEY!,
  appName: "my app",
});

overmindClient.initTracing({
  enableBatching: false,
  enabledProviders: { googleGenAI: GoogleGenAI },
});

const client = new GoogleGenAI.GoogleGenAI({
  apiKey: process.env.GEMINI_API_KEY,
});

const response = await client.models.generateContent({
  model: "gemini-2.0-flash",
  contents: "Explain quantum computing",
});

console.log(response.text);

await overmindClient.shutdown();

That’s it — quick setup and your existing LLM code works as-is. No proxy, no key sharing.

Traces are forwarded to https://api.overmindlab.ai and appear in your Overmind dashboard.

---

4. Send 30+ Traces

Run your application normally. Once Overmind has collected at least 30 traces, the optimization engine kicks in:

• Prompt templates are automatically extracted from your traces, creating Agents
• The LLM judge evaluates each trace on quality, cost, and latency
• Prompt and model experimentation begins

You can watch traces appear in real time at console.overmindlab.ai under the Traces tab.

---

5. Check the Home page

Go to the Overmind home page. You’ll see:

• Agents: We extract all prompt templates from your traces and automatically create Agents. Each agent has its own prompt template, evaluation criteria and model backtesting.
• Suggestions: For each agent we will automatically create Suggestions - a result of our optimisation engine which will suggest a better prompt or model based on each Agent’s use case.

---

What Happens Next

Once you have traces flowing, Overmind automatically:

1. Extracts your Agent — identifies the structure of your prompts and separates fixed templates from variable inputs creating Agents
2. Evaluates with your Agent — based on Agent traces we create evaluation criteria and use LLM judge to assess your Agent’s performance. We also monitor latency and cost.
3. Experiments with prompts — generates candidate prompt variations and tests them against your historical inputs on quality, cost and latency
4. Experiments with models — replays your traces through different models to find cost/quality tradeoffs
5. Shows recommendations — surfaces the best alternatives in your dashboard
6. Collects your feedback - at any time you can adjust evaluation criteria, agent description and provide feedback on our scoring and suggestions.

You provide feedback (accept, reject, tweak) and the system learns. See How Optimization Works for the full details.

---

Environment Variables

Variable	Required	Description
OVERMIND_API_KEY	Yes	Your Overmind API key (shown after signup)
OVERMIND_SERVICE_NAME	No	Service name (overrides service_name param default)
OVERMIND_ENVIRONMENT	No	Environment name (overrides environment param default)
OPENAI_API_KEY	If using OpenAI	Your OpenAI API key
ANTHROPIC_API_KEY	If using Anthropic	Your Anthropic API key
GEMINI_API_KEY	If using Google	Your Google/Gemini API key

---

Next Steps

• How Optimization Works — understand the evaluation and recommendation loop
• Python SDK Reference — full Python client API, tracing, and advanced usage
• JavaScript SDK Reference — full JS/TS client API, initTracing options, and resource attributes