feat(contrib/mcp-go): Initial mcp-go tracer implementation

[jboolean]

What does this PR do?

Adds contrib library for mcp-go. Uses it to record LLMObs Tool spans for tool calls.

This is the base of a Graphite stack.

Motivation

Tech spec for this project

Closes MLOB-4372

Reviewer's Checklist

• Changed code has unit tests for its functionality at or near 100% coverage.
• System-Tests covering this feature have been added and enabled with the va.b.c-dev version tag.
• There is a benchmark for any new code, or changes to existing code.
• If this interacts with the agent in a new way, a system test has been added.
• New code is free of linting errors. You can check this by running ./scripts/lint.sh locally.
• Add an appropriate team label so this PR gets put in the right place for the release notes.
• Non-trivial go.mod changes, e.g. adding new modules, are reviewed by @DataDog/dd-trace-go-guild.

Testing

In addition to automated tests, this was tested locally with Datadog MCP Server

Unsure? Have a question? Request a review!

[jboolean]

• feat(contrib/mcp-go): Record MCP Session ID as tag #4106
• feat(contrib/mcp-go): Trace MCP session initializations with MLObs spans #4101
• feat(contrib/mcp-go): Initial mcp-go tracer implementation #4100 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-11-05 20:59:25

Comparing candidate commit b00ba8a in PR branch jb/contrib-mcp-go-base with baseline commit 53b84e7 in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 24 metrics, 0 unstable metrics.

[datadog-official]

✅ Tests

🎉 All green!

❄️ No new flaky tests detected
🧪 All tests passed

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: b00ba8a | Docs | Datadog PR Page | Was this helpful? Give us feedback!}

[jboolean]

The in-memory cache just stores data between the start and end hooks. Creating a middleware would have avoided this, but would have been harder to automatically add to all tools. We'll also use hooks for session initialization spans (another PR on this stack).

Unfortunately this cache implementation does need to be shut down manually.

I experimented with but decided against cache customization options (max size, ttl).