feat(contrib/mark3labs/mcp-go): skip telemetry injection for UI-only tools

[jboolean]

This PR stack ports functionality added in the clone of this contrib in dd-source back to dd-trace-go.

Summary

The intent-capture ListTools hook unconditionally adds telemetry as a required field to every tool's input schema. That breaks tools annotated as UI-only (_meta.ui.visibility excluding "model"): they are invoked by the app UI, not by the model, so the UI caller has no intent to supply. OpenAI's MCP client strictly validates tool arguments against the advertised schema and rejects every UI-driven call as a result.

This change adds an isModelCallable helper that inspects *mcp.Meta.AdditionalFields["ui"]["visibility"] (handling both []string and []any shapes from JSON round-trips) and short-circuits the injection when the model is not in the visibility list. Tools with no _meta.ui.visibility continue to default to model-callable per the MCP Apps spec.

Ported from internal dd-source PR #420911.

Test plan

• TestIntentCaptureSkipsUIOnlyTools — three tools (model-only, UI-only [app], dual [app, model]); asserts telemetry is injected for model and dual but not for UI-only.
• Existing intent-capture tests still pass.

Stacked on #4941.

MCP standard reference

The _meta.ui.visibility field is part of the official MCP Apps extension. Values: ["model", "app"] (default), ["model"], or ["app"] — they control which actor may invoke the tool. UI-only tools (["app"]) are invoked by the host shell, never by the model. This PR's skip logic respects that standard.

---

Parallel port in the modelcontextprotocol/go-sdk contrib: #4949

[jboolean]

• feat(contrib/mark3labs/mcp-go): support redacting tool output in LLMObs traces #4945
• feat(contrib/mark3labs/mcp-go): support per-request intent capture #4944
• feat(contrib/mark3labs/mcp-go): set session ID on tool spans at start time #4943
• feat(contrib/mark3labs/mcp-go): skip telemetry injection for UI-only tools #4942 👈 (View in Graphite)
• feat(contrib/mark3labs/mcp-go): expose captured intent to tool handlers via context #4941
• fix(contrib/mark3labs/mcp-go): preserve raw input schema fields on intent injection #4946
• fix(contrib/mark3labs/mcp-go): avoid map race when normalizing RawInputSchema #4940
• feat(contrib/mark3labs/mcp-go): support intent capture for tools using RawInputSchema #4939
• main

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

[datadog-prod-us1-3]

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

🎯 Code Coverage (details)
• Patch Coverage: 21.74%
• Overall Coverage: 62.78% (+11.47%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 2ff4a7b | Docs | Datadog PR Page | Give us feedback!}

[pr-commenter]

Benchmarks

Benchmark execution time: 2026-07-01 16:38:29

Comparing candidate commit 2ff4a7b in PR branch jb/mcp-skip-ui-only-tools with baseline commit 2de53a9 in branch jb/mcp-intent-from-context.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 321 metrics, 2 unstable metrics, 1 flaky benchmarks without significant changes.

Explanation

This is an A/B test comparing a candidate commit's performance against that of a baseline commit. Performance changes are noted in the tables below as:

• 🟩 = significantly better candidate vs. baseline
• 🟥 = significantly worse candidate vs. baseline

We compute a confidence interval (CI) over the relative difference of means between metrics from the candidate and baseline commits, considering the baseline as the reference.

If the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD), the change is considered significant.

Feel free to reach out to #apm-benchmarking-platform on Slack if you have any questions.

More details about the CI and significant changes

You can imagine this CI as a range of values that is likely to contain the true difference of means between the candidate and baseline commits.

CIs of the difference of means are often centered around 0%, because often changes are not that big:

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

As described above, a change is considered significant if the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD).

For instance, for an execution time metric, this confidence interval indicates a significantly worse performance:

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

Known flaky benchmarks

These benchmarks are marked as flaky and will not trigger a failure. Modify FLAKY_BENCHMARKS_REGEX to control which benchmarks are marked as flaky.

Known flaky benchmarks without significant changes:

• scenario:BenchmarkOTLPTraceWriterFlush

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a8261f9663

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Handle typed UI metadata before defaulting callable

When a server constructs the same valid _meta.ui.visibility wire format with a typed Go value, such as a struct or map[string][]string, this assertion falls through and treats the tool as model-callable. mcp.Meta.AdditionalFields accepts arbitrary values and will marshal those typed values correctly, so app-only tools declared that way still get required telemetry injected in tools/list, leaving the UI calls rejected for the scenario this change is meant to fix.

Useful? React with 👍 / 👎.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (jb/mcp-intent-from-context@0c444d5). Learn more about missing BASE report.

Additional details and impacted files

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.