Agent Cost & UsageSupported Agentic Stacks

Supported Agentic Stacks

Use this page to check whether the technologies used by the agent application are supported by the Mavvrik SDK.

The tables below show support by SDK language. Support for third-party libraries can depend on the installed library version; after upgrading a provider or framework dependency, confirm that instrumentation still loads successfully.

Languages and runtimes

Language / runtime	Support
Python	Supported
JavaScript / TypeScript	Supported
Node.js	Supported
Bun	Supported
Deno	Supported

For Python applications, install the supported Mavvrik 1.3.x SDK series:

Bash

pip install "mvk-sdk-py[genai]~=1.3.0"

JavaScript / TypeScript applications use:

Bash

npm install @mavvrikai/sdk

and the required @mvk/instr-* packages.

LLM providers and model platforms

Integration	Python	JavaScript / TypeScript
OpenAI	Supported	Supported
Azure OpenAI	Supported	Supported
Anthropic	Supported	Supported
Anthropic on Vertex	Supported	Supported
Amazon Bedrock	Supported	Supported
Amazon Bedrock AgentCore	Supported	Supported
Google Gemini	Supported	Supported
Google GenAI	Supported	Supported
Google Generative Language	Supported	Not supported
Google Vertex AI	Supported	Supported
Perplexity	Supported	Supported
Oracle OCI Generative AI	Supported	Not supported
OpenRouter	Supported	Not supported
LiteLLM library	Supported	Not supported

LiteLLM library in this table means Python SDK instrumentation of the LiteLLM library. It is different from the separate LiteLLM Proxy integration, which sends gateway telemetry to Mavvrik.

Agent frameworks

Framework	Python	JavaScript / TypeScript
LangChain	Supported	Supported
LangGraph	Supported	Supported
CrewAI	Supported	Not supported
Agno	Supported	Not supported
Semantic Kernel	Supported	Not supported
OpenAI Agents	Supported	Supported
Microsoft Agent Framework	Supported	Not supported

For eve, use the dedicated Connect eve integration. eve sends its native OpenTelemetry telemetry directly to Mavvrik and does not require the Mavvrik SDK.

Vector databases

Integration	Python	JavaScript / TypeScript
Pinecone	Supported	Supported
ChromaDB	Supported	Supported
Qdrant	Supported	Supported
Weaviate	Supported	Supported
Milvus	Not supported	Supported
pgvector	Not supported	Supported

Web frameworks and serverless runtimes

Runtime / framework	Python	JavaScript / TypeScript
FastAPI / Starlette	Supported	—
Flask	Supported	—
Django	Supported	—
Generic ASGI / WSGI	Supported	—
Express	—	Supported
Fastify	—	Supported
Koa	—	Supported
Hono	—	Supported
NestJS	—	Supported
Next.js	—	Supported
AWS Lambda	Supported	Supported
Google Cloud Functions / Cloud Run	Supported; flush may be required	Supported
Azure Functions	Supported; flush may be required	Supported

HTTP instrumentation

HTTP instrumentation is not enabled by default. Python supports explicit httpx instrumentation. Request and response bodies are not read.

Bash

export MVK_HTTP_EXCLUSIONS='["internal-billing.corp","patient-records.internal"]'

Do not enable HTTP instrumentation when AI-provider or framework instrumentation already supplies the required cost visibility.

Python installation groups

Bundle	Includes
`[genai]`	Supported AI providers and related integrations
`[frameworks]`	Supported agent frameworks
`[vectordb]`	Supported vector databases
`[all]`	All available Python integrations

Individual extras can be installed when only specific integrations are required:

Bash

pip install "mvk-sdk-py[openai]~=1.3.0"
pip install "mvk-sdk-py[openai,anthropic,langchain]~=1.3.0"

JavaScript / TypeScript instrumentation packages

Bash

npm install @mavvrikai/sdk @mvk/instr-openai
npm install @mavvrikai/sdk @mvk/instr-anthropic
npm install @mavvrikai/sdk @mvk/instr-langchain

Third-party library versions

Mavvrik integrations recognize supported versions of third-party provider and framework libraries. If a dependency version is not recognized by the installed SDK, that integration can be skipped and logged while the application continues to run.

After upgrading a provider or framework dependency, enable DEBUG logging and confirm that the integration loads successfully:

Bash

export MVK_LOG_LEVEL=DEBUG

Verify a supported integration

Install the required Mavvrik package or extra.
Initialize Mavvrik before the provider or framework is loaded.
Run one known operation.
Confirm the integration appears in DEBUG startup output.
Confirm the operation appears in Home → Agentic → Cost and Home → Agentic → Sessions.