Integrations

LLM Providers

Python

Provider	Status	providers value
OpenAI	Supported	"openai"
Anthropic	Supported	"anthropic"
Google Gemini	Supported	"google"
Agno	Supported	"agno"

Call overmind_sdk.init() once at startup with the providers you use. Your existing LLM client code stays unchanged — no import swaps, no proxy routing.

from overmind_sdk import init

init(service_name="my-service", providers=["openai", "anthropic"])

Pass an empty providers list (or omit it) to auto-detect all installed providers.

JavaScript / TypeScript

Provider	Status	enabledProviders key	Package
OpenAI	Supported	openai: OpenAI	openai
Anthropic	Supported	anthropic: Anthropic	@anthropic-ai/sdk
Google Gemini	Supported	googleGenAI: GoogleGenAI	@google/genai

Pass the imported provider module to initTracing({ enabledProviders: { ... } }). Your existing LLM code works unchanged.

---

Frameworks

Framework	Status	Notes
Agno	Supported	Traces agent runs, tool calls, and reasoning steps
OpenAI Agents SDK	Supported	Traces agent tool calls and reasoning steps
LangChain	Coming soon	Protocol-level support for chains and agents
CrewAI	Coming soon	Multi-agent workflow tracing

---

Auto-Instrumentation

Python

The Python SDK uses OpenTelemetry auto-instrumentation. Call init() once at startup and every subsequent LLM call from your existing client is automatically traced.

• OpenAI — captures all chat completion, response, and embedding calls
• Anthropic — captures all message calls
• Google Gemini — captures all generate_content calls
• Agno — captures agent runs, tool calls, and model interactions

from overmind_sdk import init

init(service_name="my-service")  # instruments all installed providers

JavaScript / TypeScript

The JS SDK instruments LLM calls via initTracing(). Call it once before any LLM calls and all subsequent provider calls are traced automatically.

• OpenAI — captures chat completions, completions, and image generation calls
• Anthropic — captures all message calls
• Google Gemini — captures all generateContent and streaming calls

---

Sending Traces from Any Language

If you’re not using Python, you can send traces directly via the OTLP HTTP endpoint:

POST https://api.overmindlab.ai/api/v1/traces
Header: X-API-Token: ovr_your_token_here
Content-Type: application/x-protobuf

The endpoint accepts standard OpenTelemetry OTLP trace data. Any OpenTelemetry SDK (Go, Java, Node.js, etc.) can export to this endpoint.

---

What’s Next

We’re actively adding support for more providers and frameworks. If you have a specific integration request, reach out to support@overmindlab.ai