Provider Tools

Most providers expose native tools beyond user-defined function calls: web search, code execution, computer use, hosted retrieval, and more. TanStack AI exports each provider's native tools from a dedicated /tools subpath per adapter package.

You have an adapter already wired up. You want to give the model access to a provider-native capability (e.g. Anthropic web search) and be sure you never pair a tool with a model that doesn't support it. By the end of this page, you'll have imported the factory, added it to chat({ tools: [...] }), and understood the compile-time guard that will catch unsupported combinations.

Import#

Every adapter ships provider tools on a /tools subpath:

import { webSearchTool } from '@tanstack/ai-anthropic/tools'
import { codeInterpreterTool } from '@tanstack/ai-openai/tools'
import { googleSearchTool } from '@tanstack/ai-gemini/tools'

Use in chat({ tools })#

import { chat } from '@tanstack/ai'
import { anthropicText } from '@tanstack/ai-anthropic'
import { webSearchTool } from '@tanstack/ai-anthropic/tools'

const stream = chat({
  adapter: anthropicText('claude-opus-4-6'),
  messages: [{ role: 'user', content: "Summarize today's AI news." }],
  tools: [
    webSearchTool({
      name: 'web_search',
      type: 'web_search_20250305',
      max_uses: 3,
    }),
  ],
})

Multi-turn persistence#

Provider tools run on the provider's own infrastructure, so their results (e.g. Anthropic web_search sources, web_fetch page contents) come back embedded in the assistant turn rather than as a separate tool message. TanStack AI preserves those results on the assistant message, so when you feed the prior conversation back into the next chat() call the model still sees the earlier evidence — no special handling required:

import { chat, StreamProcessor } from '@tanstack/ai'
import { anthropicText } from '@tanstack/ai-anthropic'
import { webSearchTool } from '@tanstack/ai-anthropic/tools'

const adapter = anthropicText('claude-opus-4-6')
const tools = [webSearchTool({ name: 'web_search', type: 'web_search_20250305' })]

const processor = new StreamProcessor()
for await (const chunk of chat({
  adapter,
  tools,
  messages: [{ role: 'user', content: 'Find two sources on the drone market.' }],
})) {
  processor.processChunk(chunk)
}
processor.finalizeStream()

// The follow-up turn can still cite the previous search results.
const followUp = chat({
  adapter,
  tools,
  messages: [
    ...processor.getMessages(),
    { role: 'user', content: 'List the exact sources you used.' },
  ],
})

The search/fetch call surfaces as a provider-executed tool-call part on the assistant message; the agent loop never tries to run it client-side.

Type-level guard#

Every provider-specific tool factory (e.g. webSearchTool, computerUseTool) returns a ProviderTool<TProvider, TKind> brand. The adapter's toolCapabilities (derived from each model's supports.tools list) gates which brands are assignable to tools.

Paste a computerUseTool(...) into a model that doesn't expose it, and TypeScript reports an error on that array element — not on the factory call, not at runtime. User-defined toolDefinition() tools stay unbranded and always assignable. The customTool factories exported from ai-anthropic and ai-openai also return a plain Tool (not a ProviderTool brand) and are therefore universally accepted by any chat model, just like toolDefinition().

Available tools#

Which models support which tools?#

Each adapter's supports.tools array is the source of truth. The comparison matrix is maintained alongside model-meta.ts and reflected here:

• Anthropic: every registered model supports the full tool superset (the retired Claude 3.x models with narrower support were removed).

• OpenAI: GPT-5 family and reasoning models (O-series) support the full superset. GPT-4-series supports web/file/image/code/mcp but not preview/shell variants. GPT-3.5 and audio-focused models: none.

• Gemini: 3.x Pro/Flash models support the full tool set. Lite and image/video variants have narrower support.

• OpenRouter: every chat model supports webSearchTool and webFetchTool via the gateway.

  For the exact per-model list, open the adapter page or read the model's supports.tools array directly from model-meta.ts.

Provider Skills#

Anthropic and OpenAI support hosted, provider-managed skill bundles that run inside the provider's server-side sandbox. Skills attach to an execution tool (codeExecutionTool for Anthropic, shellTool for OpenAI) and are referenced by ID — the provider handles installation and execution.

See Provider Skills for full setup steps and examples.

Migrating from earlier versions#

If you were using createWebSearchTool from @tanstack/ai-openrouter, see Migration Guide §6.