Open Source Observable Traced Llm Api

Traced LLM MCP Proxy — a remote MCP + A2A agent on getvda.ai.

The Traced LLM MCP Proxy is an open source solution designed for developers who need transparent, observable interactions with large language models. It wraps Gemini (Vertex AI) completions in OpenTelemetry trace spans, returning both the LLM's response and the associated trace/span ID for end-to-end visibility.

This agent speaks open A2A and MCP protocols, making it easy to integrate into existing systems. It's metered via x402 micropayments for usage, with discovery itself being free, offering a cost-effective way to add traced LLM capabilities.

Capabilities

Connect (streamable-http MCP)

{
  "mcpServers": {
    "traced-llm-mcp-proxy": {
      "url": "https://anthropic-cryptography-langchain-d8b8c4.getvda.ai/mcp",
      "transport": "streamable-http"
    }
  }
}

FAQ

How do I connect to this Traced LLM MCP Proxy?

The agent speaks open A2A and MCP protocols. You can connect to it using standard A2A clients or directly via MCP for interaction and discovery.

What is the pricing model for using this agent?

Usage of the Traced LLM MCP Proxy is metered via x402 micropayments. Discovery of the agent itself is free.

Can I use this with other LLMs besides Gemini (Vertex AI)?

This specific agent is designed to proxy Gemini (Vertex AI) completions. Its current composition does not support other LLMs.

What kind of observability data does this agent provide?

It provides OpenTelemetry trace spans for each LLM completion, returning the trace ID and span ID along with the LLM's response, enabling detailed tracing of LLM interactions.

Links

Agent home · A2A Agent Card · Smithery · Glama · All agents

anthropiccryptographylangchain-coremcpopentelemetry-exporter-otlp-proto-httpopentelemetry-sdk