fix(openrouter): disable Claude Code tool search for OpenRouter sessions

[gfsaaser24]

Any OpenRouter turn on a non-Anthropic model failed with:

400 Deferred custom tools are only supported on Anthropic models.
Non-Anthropic models cannot call tools omitted from tools[].
Received moonshotai/kimi-k3-20260715.

Claude Code defaults to tool search: it omits tools from tools[] and refers to them with tool_reference blocks. OpenRouter's Anthropic-compatible endpoint rejects that for every non-Anthropic model.

Claude Code does have an auto-disable for non-first-party base URLs, but reading the shipped CLI shows it only fires when the resolved auth kind is firstParty:

if (!K.ENABLE_TOOL_SEARCH && !l3o() && Yn() === "firstParty" && !qf()) return false

Once ANTHROPIC_AUTH_TOKEN carries the OpenRouter key that is no longer the case, so the guard never triggered and tool search stayed on.

Fix: set ENABLE_TOOL_SEARCH=false in the OpenRouter process env. That is Claude Code's own off switch — its parser maps false to "standard" mode, which returns false from the tool-search gate. An explicit host value still wins, for proxies that do forward tool_reference blocks; a blank host value falls back to the OpenRouter default.

Tests cover both the default and the host override.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: f3d5d10e-7aee-4799-a5d9-efdf612bcbc5

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	12.6 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	6.3 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.3 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	51.8 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	16	—	21	✅
Claude	Total thread wire	—	12.6 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	6.3 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.3 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	52.7 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	16	—	21	✅

Baseline: unavailable · PR result: f9ff460 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 99.9 KiB
• Claude decoded thread snapshot: 100.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.