Registry / observability / openinference-instrumentation-mcp

openinference-instrumentation-mcp

library2.0.3pypypiunverified

OpenInference instrumentation for MCP (Model Context Protocol) clients, enabling OpenTelemetry-based tracing of MCP tool calls and server interactions. Current version 2.0.0, requires Python >=3.10 <3.15. Part of the Arize AI OpenInference ecosystem. Active development with frequent releases.

observability llm-agents http-networking

pip install openinference-instrumentation-mcp

github ↗package ↗

INSTALL

IMPORT

SIG · OPENINFERENCE-INST

openinference-instrumentation-mcp

observabilitypythonv2.0.3

Install

2.6s avg

Import

—

Disk

22MB

Pass rate

8/ 10

Env Coverage8 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v2.0.3 · pip install

no network on importno background threads

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

glibc

py 3.10

✓ —

✓ 2.9s

py 3.11

✓ —

✓ 2.7s

py 3.12

✓ —

✓ 2.5s

py 3.13

✓ —

✓ 2.4s

py 3.9

✕ build_error

22MB installed

● package 22MB

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

MCPClientInstrumentor

✓ from openinference.instrumentation.mcp import MCPClientInstrumentor

✗ from openinference_instrumentation_mcp import MCPClientInstrumentor

Instrument MCP client with OpenInference tracing.

import os from opentelemetry import trace from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter from opentelemetry.sdk.trace import TracerProvider from opentelemetry.sdk.trace.export import SimpleSpanProcessor from openinference_instrumentation_mcp import MCPClientInstrumentor # Set up OpenTelemetry tracer provider provider = TracerProvider() provider.add_span_processor(SimpleSpanProcessor(OTLPSpanExporter(endpoint=os.environ.get('OTEL_EXPORTER_OTLP_ENDPOINT', 'http://localhost:4318/v1/traces')))) trace.set_tracer_provider(provider) # Instrument MCP client MCPClientInstrumentor().instrument() # Now any MCP client sessions will be traced.

Debug

Known issues

breakingVersion 2.0.0 introduces a BREAKING CHANGE: support for streamable_http_client. Existing instrumentations using older MCP client versions may break. Update MCP library to compatible version.

fix

Upgrade MCP library to version that supports streamable_http_client (mcp>=1.6.0). Review your MCP server's streaming transport.

affects: <2.0.0

gotchaInstrumentation must be set up before creating any MCP client sessions. If MCP client is imported and used before calling MCPClientInstrumentor().instrument(), the instrumentation will not apply.

fix

Call MCPClientInstrumentor().instrument() immediately after setting up the tracer provider and before importing/using mcp.ClientSession or any MCP client.

affects: all

gotchaThe instrumentation only instruments the client side, not the server. Do not expect server-side spans unless you set up instrumentation on the server as well.

fix

Use a separate OpenInference instrumentation for your MCP server (e.g., openinference-instrumentation-mcp-server) if available.

affects: all

deprecatedWrapt 2.x is supported, but older instrumentations may require wrapt<2. If you encounter import errors related to wrapt, check compatibility.

fix

Ensure wrapt is installed. For wrapt 2.x compatibility, use openinference-instrumentation>=0.1.47.

affects: >=1.0.0

Upgrade

Version history

2.0.3latest on PyPI · released May 18, 2026

Audit

Dependencies

openinference-instrumentationrequiredCore instrumentation base classes and utilities

mcprequiredRequired to instrument MCP client sessions

openinference-semantic-conventionsrequiredSemantic convention constants for OpenInference spans

opentelemetry-apirequiredOpenTelemetry API for span creation and context propagation

wraptrequiredUsed for monkey-patching MCP client methods

Agent activity

34 hits · last 30 days

node

OpenAI (training)

Amazon

Resources

githubhttps://github.com/Arize-ai/openinference ↗

packagehttps://pypi.org/project/openinference-instrumentation-mcp/ ↗