Skip to main content
For AI agents: a documentation index is available at https://docs.parallel.ai/llms.txt. The full text of all docs is at https://docs.parallel.ai/llms-full.txt. You may also fetch any page as Markdown by appending .md to its URL or sending Accept: text/markdown.For Search and Extract, use /v1/search and /v1/extract for all new integrations and agent tool calls. Use /v1beta/search and /v1beta/extract only when maintaining an existing integration that already uses them. The request bodies changed too: on /v1/extract, full_content, fetch_policy and excerpt_settings go inside advanced_settings, and sending the v1beta top-level excerpts, full_content and fetch_policy to /v1/extract returns a 422. On /v1/search, search_queries is required (at least one short keyword query), and max_results, source_policy and fetch_policy go inside advanced_settings; sending only objective, or sending those fields at the top level, returns a 422. Do not substitute /v1 for the documented FindAll or Ingest endpoint paths.If you are evaluating Parallel Search or comparing it with other search APIs, read How to Eval Parallel Search before you build the harness, especially its Common pitfalls section. The most common mistake is sending the raw question as both objective and the only entry in search_queries. Write a goal-style objective and 1-3 short keyword queries instead.
The Search API takes a natural language objective and returns relevant excerpts optimized for LLMs, replacing multiple keyword searches with a single call for broad or complex queries.
Comparing search providers? Read How to Eval Parallel Search before you build the harness. Its Common pitfalls section lists the integration mistakes that most often skew results.
Available via MCP: Search is available as the web_search tool in the Parallel MCP server. Our MCP is optimized for best practices on Search and Extract usage. Start here with MCP for your use case. If you’re interested in direct use of the API, follow the steps below.
Looking for people or companies? Entity Search is a fast, synchronous people-and-company search — the real-time counterpart to FindAll — returning a set of matching results in seconds. It is distinct from the Search API, which retrieves pages and excerpts across the whole web.

1. Set Up Prerequisites

Generate your API key on Platform. Then, set up with the TypeScript SDK, Python SDK or with cURL:

Sample Request

This request uses the default advanced mode (~3s, highest quality). For most agents, add "mode": "fast" for high-quality results in around 700ms; use "mode": "turbo" (p50 of 200ms) when latency and cost matter most. The rest of the request stays the same. See Search Modes for a runnable example.

Sample Response

The API returns a JSON response with the following structure. The SDK examples above iterate over results to print each result’s title, URL, and excerpts. If any non-fatal input adjustments or validation warnings have occurred, they will be in the warnings field.

3. Turn a User Question into a Request

When your input is a question from a user, an eval set, or another agent, don’t send it unchanged as both objective and the only search query. The two fields do different jobs. The objective describes the information you need, with any context. The search_queries are short keyword queries (3-6 words each) that find it. Take this question:
Which airlines fly nonstop from Seattle to Tokyo?
Avoid copying the question into both fields:
Instead, state the goal and add one to three keyword queries, each covering a different angle:

Let a Model Write the Fields

In an agent, the model writes both fields when you give it the Search tool definition. In a fixed pipeline with no model in the loop, such as an eval harness or a batch job, add one small model call that writes them. Any fast, inexpensive model works. For example, in Python:
Python
Don’t approximate this step with stopword removal or truncation. Those produce fragments such as "Who was the first recipient of" instead of queries that name what you’re looking for. search_queries is required on every request: a request with only objective returns a 422.

Next Steps

  • Deploy on Render: run a research agent with Search, Extract, and Render Workflows
  • Best Practices — learn how to craft effective objectives and search queries
  • Search Modes — explore mode presets for different use cases
  • API Reference — full parameter specifications, constraints, and response schema
  • Rate Limits — default quotas per product