--- title: Your AI Tool Calls Should Fail at Compile Time, Not in... description: TanStack AI's latest release adds compile-time type gating for provider-hosted tools like Anthropic web search, OpenAI code interpreter, and Gemini Google... canonical: https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Your AI Tool Calls Should Fail at Compile Time, Not in Production | daily.dev og:description: TanStack AI's latest release adds compile-time type gating for provider-hosted tools like Anthropic web search, OpenAI code interpreter, and Gemini Google... og:url: https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip og:image: https://api.daily.dev/og/posts/osTnYiWIp.png og:image:alt: Your AI Tool Calls Should Fail at Compile Time, Not in Production og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Your AI Tool Calls Should Fail at Compile Time, Not in Production **[TanStack](https://daily.dev/sources/tanstack-blog)** · 6 min read · 0 upvotes · 0 comments ## Summary TanStack AI's latest release adds compile-time type gating for provider-hosted tools like Anthropic web search, OpenAI code interpreter, and Gemini Google search. Because tool support varies per model (not per provider), passing an unsupported tool previously failed silently at runtime, with the model ignoring the tool and generating a plausible-but-wrong answer. Now every provider tool factory returns a phantom-branded ProviderTool type checked against each model's capability list, so mismatched pairings produce a TypeScript compile error at the exact line where the tool is passed. User-defined tools remain universally assignable and unaffected. Tools are also now imported from per-adapter /tools subpaths for better tree-shaking. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why does my Anthropic web search or computer use tool silently fail instead of throwing an error? Provider-hosted tools like Anthropic's computer use or web search are only supported on specific models, not across an entire provider's lineup. If you pass an unsupported tool, the provider either rejects the request or, more commonly, the model quietly ignores the tool and generates a confident but wrong answer with no error or stack trace. For example, claude-3-haiku only supports web_search, while claude-opus-4-6 supports the full tool superset including computer use and bash. _Developers pairing LLM tools with specific models can track SDK safety improvements like this via daily.dev._ ### How does TanStack AI prevent passing an incompatible provider tool to a model at compile time? TanStack AI now brands every provider tool factory with a ProviderTool type, and each adapter carries a toolCapabilities list derived from the model's supports.tools array. TypeScript checks the tools array against that capability list, so passing an unsupported tool like computerUseTool to claude-3-haiku produces a compile error on that exact line instead of failing silently in production. The brand is erased at runtime with zero bundle or performance cost. _Teams building type-safe AI integrations can follow tooling changes like this through daily.dev._ ### Do I need to change how I use custom, user-defined tools with TanStack AI's new provider tool type gating? No, custom tools created with toolDefinition() or the customTool() factories from OpenAI and Anthropic adapters remain plain, unbranded Tool objects that are universally assignable to any model. Only provider-hosted tools such as web search or code interpreter are gated by model capability, since those are the only tools with per-model support differences; self-written tools run wherever your code runs. _daily.dev helps developers stay current on SDK changes that affect existing AI tool integrations._ ## Similar posts on daily.dev - [A Guide to Tool Calling with the TypeScript AI SDK](https://daily.dev/posts/a-guide-to-tool-calling-with-the-typescript-ai-sdk-f11xvpxtm) · Telerik · 0 upvotes · 0 comments - [TanStack Releases Framework Agnostic AI Toolkit](https://daily.dev/posts/tanstack-releases-framework-agnostic-ai-toolkit-j747yitl6) · InfoQ · 1 upvotes · 0 comments - [Your Agent Isn't Dumb. Your Tools Are.](https://daily.dev/posts/your-agent-isn-t-dumb-your-tools-are--grledaace) · The T-Shaped Dev · 3 upvotes · 2 comments - [Agent-as-a-Tool: A New Era of AI Orchestration](https://daily.dev/posts/agent-as-a-tool-a-new-era-of-ai-orchestration-vxph42s1s) · DEV · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#typescript](https://daily.dev/tags/typescript), [#tanstack](https://daily.dev/tags/tanstack) [View this post on daily.dev](https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Your AI Tool Calls Should Fail at Compile Time, Not in Production","url":"https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip"},"datePublished":"2026-08-23T12:22:22.171Z","dateModified":"2026-08-23T12:49:03.754Z","description":"TanStack AI's latest release adds compile-time type gating for provider-hosted tools like Anthropic web search, OpenAI code interpreter, and Gemini Google...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/697a2fb451718ff5711750927cd8ff84?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/697a2fb451718ff5711750927cd8ff84?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"TanStack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"TanStack","logo":"https://media.daily.dev/image/upload/s--TVIrvIT7--/f_auto,q_auto/v1787487634/logos/tanstack-blog?_a=BAMAMicg0","url":"https://daily.dev/sources/tanstack-blog"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,typescript,tanstack","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"TanStack","item":"https://daily.dev/sources/tanstack-blog"},{"@type":"ListItem","position":3,"name":"Your AI Tool Calls Should Fail at Compile Time, Not in Production"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/your-ai-tool-calls-should-fail-at-compile-time-not-in-production-ostnyiwip#faq","mainEntity":[{"@type":"Question","name":"Why does my Anthropic web search or computer use tool silently fail instead of throwing an error?","acceptedAnswer":{"@type":"Answer","text":"Provider-hosted tools like Anthropic's computer use or web search are only supported on specific models, not across an entire provider's lineup. If you pass an unsupported tool, the provider either rejects the request or, more commonly, the model quietly ignores the tool and generates a confident but wrong answer with no error or stack trace. For example, claude-3-haiku only supports web_search, while claude-opus-4-6 supports the full tool superset including computer use and bash. Developers pairing LLM tools with specific models can track SDK safety improvements like this via daily.dev."}},{"@type":"Question","name":"How does TanStack AI prevent passing an incompatible provider tool to a model at compile time?","acceptedAnswer":{"@type":"Answer","text":"TanStack AI now brands every provider tool factory with a ProviderTool type, and each adapter carries a toolCapabilities list derived from the model's supports.tools array. TypeScript checks the tools array against that capability list, so passing an unsupported tool like computerUseTool to claude-3-haiku produces a compile error on that exact line instead of failing silently in production. The brand is erased at runtime with zero bundle or performance cost. Teams building type-safe AI integrations can follow tooling changes like this through daily.dev."}},{"@type":"Question","name":"Do I need to change how I use custom, user-defined tools with TanStack AI's new provider tool type gating?","acceptedAnswer":{"@type":"Answer","text":"No, custom tools created with toolDefinition() or the customTool() factories from OpenAI and Anthropic adapters remain plain, unbranded Tool objects that are universally assignable to any model. Only provider-hosted tools such as web search or code interpreter are gated by model capability, since those are the only tools with per-model support differences; self-written tools run wherever your code runs. daily.dev helps developers stay current on SDK changes that affect existing AI tool integrations."}}]} ```