Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.agnost.ai/llms.txt

Use this file to discover all available pages before exploring further.

This page gets one tool invocation or AI interaction visible in the Agnost dashboard. Pick the path that matches what you’re building.

​
1. Get your organization ID

  1. Sign in at app.agnost.ai.
  2. Open Settings → Organization. Copy your organization ID: it’s a UUID.
This ID is what we’ll refer to as your org_id (also called write_key in some SDKs). Treat it as a public identifier: it does not grant write access on its own.

​
2. Pick a path

MCP server

Track tool invocations on a Model Context Protocol server. Python, TypeScript, or Go.

AI conversations

Track LLM calls / agent turns from any Python or Node application.

OpenTelemetry agent

You’re already running Vercel AI SDK, Mastra, OpenAI Agents, LangChain, etc. and want to forward traces.

MCP Toolbox

Google’s genai-toolbox for databases. One-flag setup.

​
Path A: MCP server

pip install agnost-mcp

from mcp.server.fastmcp import FastMCP
from agnost_mcp import track

server = FastMCP("my-server")

@server.tool()
def hello(name: str) -> str:
    return f"Hello, {name}!"

track(server, "your-org-id")

if __name__ == "__main__":
    server.run()
See the FastMCP guide for options.
Invoke a tool from your MCP client. Within a few seconds the call appears under Tools in the dashboard.

​
Path B: AI conversations

For tracking LLM calls or agent turns directly from your application code (no MCP). 
pip install agnost

import agnost

agnost.init("your-org-id")

interaction = agnost.begin(
    user_id="user_123",
    agent_name="my-agent",
    input="What is the weather?",
)
# ... call your model ...
interaction.end(output="Sunny and 72°F.")
See the Python Conversation SDK and configuration reference.

​
Path C: OpenTelemetry

If you’re already exporting OTel traces from a supported framework, point them at Agnost using the X-Agnost-Org-ID header: 
OTEL_EXPORTER_OTLP_ENDPOINT=https://otel.agnost.ai/v1/traces
OTEL_EXPORTER_OTLP_HEADERS=X-Agnost-Org-ID=your-org-id
Framework-specific guides: For the full integration list, see Integrations.

​
Verify

Open app.agnost.ai and check:
  • Tools: for MCP tool invocations
  • Conversations: for begin/end interactions
  • Raw logs: for the unfiltered event stream
Events typically appear within 5 seconds. Nothing showing up? See Troubleshooting.

​
Troubleshooting

  • No data after 30 seconds: confirm the org ID is correct (Settings → Organization), and that endpoint defaults to https://api.agnost.ai. For OTel, confirm the header name is X-Agnost-Org-ID (case-insensitive but the dash placement matters).
  • ModuleNotFoundError: mcp: the Python MCP SDK depends on the official mcp>=1.10 package. Install it via pip install mcp (agnost-mcp lists it as a dependency).
  • Authentication errors: org ID alone is enough for SDK ingestion. For programmatic dashboard access, see Authentication.

​
Next steps

  • SDK index: every published package and its canonical import.
  • Authentication: API keys and JWT for the REST API.
  • Errors: error response schema and status codes.
  • Architecture: how ingestion, processing, and the dashboard fit together.