Use Perplexity Web Search with Vercel AI Gateway

Link to headingHow it works

To use provider-agnostic search with the AI SDK, import gateway.tools.perplexitySearch() from @ai-sdk/gateway and pass it in the tools parameter as perplexity_search to any model. When the model determines it needs current information, it automatically searches the web and incorporates the results into its response.

1
import { streamText } from 'ai';
2
import { gateway } from '@ai-sdk/gateway';
3

4
const result = streamText({
5
  model: 'minimax/minimax-m2.1',
6
  prompt: 'What were the major tech announcements this week?',
7
  tools: {
8
    perplexity_search: gateway.tools.perplexitySearch(),
9
  },
10
});

Link to headingUse cases

Link to headingModels without native search

Many capable models don't include built-in web search. Perplexity search adds this capability to any model available through AI Gateway.

1
import { streamText } from 'ai';
2
import { gateway } from '@ai-sdk/gateway';
3

4
const result = streamText({
5
  model: 'zai/glm-4.7', // No native search, but Perplexity works
6
  prompt: 'What new features were added in React 19.2?',
7
  tools: {
8
    perplexity_search: gateway.tools.perplexitySearch(),
9
  },
10
});

Link to headingDeveloper tooling and CI assistants

Code assistants are more useful when they know about current package versions, recently merged PRs, release notes, and documentation updates. Stale information leads to broken builds and security issues.

1
import { streamText } from 'ai';
2
import { gateway } from '@ai-sdk/gateway';
3

4
const result = streamText({
5
  model: 'anthropic/claude-sonnet-4.5',
6
  prompt: 'What is the latest stable version of Next.js and what breaking changes should I know about?',
7
  tools: {
8
    perplexity_search: gateway.tools.perplexitySearch(),
9
  },
10
});

Link to headingProvider-agnostic chatbot

Production chatbots often need to work across multiple models for cost optimization, failover, or A/B testing. Native search tools differ between providers, but Perplexity search works identically regardless of which model you use.

1
import { streamText } from 'ai';
2
import { gateway } from '@ai-sdk/gateway';
3

4
const models = ['openai/gpt-5.2', 'anthropic/claude-sonnet-4.5', 'zai/glm-4.7'];
5

6
const result = streamText({
7
  model: models[selectedIndex], // Same search logic works across any model
8
  prompt: 'What is the current status of the OpenAI API?',
9
  tools: {
10
    perplexity_search: gateway.tools.perplexitySearch({
11
      searchDomainFilter: ['status.openai.com', 'openai.com'],
12
      searchRecencyFilter: 'day',
13
    }),
14
  },
15
});

Link to headingOperational and market-aware agents

Agents that monitor pricing changes, service outages, job postings, or social announcements need near real-time information to be useful. Web search keeps them current without manual data pipelines.

1
import { streamText } from 'ai';
2
import { gateway } from '@ai-sdk/gateway';
3

4
const result = streamText({
5
  model: 'openai/gpt-5.2',
6
  prompt: 'Are there any ongoing incidents affecting Vercel or AWS us-east-1?',
7
  tools: {
8
    perplexity_search: gateway.tools.perplexitySearch({
9
      searchDomainFilter: ['status.vercel.com', 'health.aws.amazon.com'],
10
      searchRecencyFilter: 'day',
11
    }),
12
  },
13
});

Link to headingProvider-specific tools

AI Gateway also supports native web search from Anthropic, OpenAI, and Google. These provider-specific tools are optimized for their respective models and may offer additional features.

Use Perplexity when you want consistent search behavior across different models, or when your model's provider doesn't offer native web search.

See the web search docs for details on all available options.

Link to headingPricing

Perplexity web search requests are charged at $5 per 1,000 requests in AI Gateway. There is no markup on these requests. See Perplexity's pricing for more details.

Link to headingGet started

Add real-time web search to your AI application today:

• Web search documentation

• AI Gateway