Truly AdaptiveSearch

Operator is an adaptive search agent that delivers faster, higher-quality answers.

Integrate In Minutes

from openai import OpenAI

client = OpenAI(
    base_url="https://api.operator.io/v1",
    api_key="op_..."
)

results = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "nextjs 15 middleware headers error"}]
)

Get an API key here ->Open Playground

Backed by Y Combinator

Agent gets stuck

I'm getting "TypeError: Cannot read properties of undefined (reading 'headers')" after upgrading to Next.js 15. The middleware was working fine in 14.

→ Agent calls Operator for context

Operator returns context

Breaking change found

Next.js 15 changed how middleware handles the request object. Headers are now accessed differently.

Sources

  • Next.js 15 Migration Guide
  • GitHub Issue #58294
  • Vercel Docs: Middleware API

Fix

Use request.headers.get() instead of request.headers

Why Operator

Fewer tokens per search

Improves speed and answer quality by cutting redundant context.

Less wasted context

Keeps simple lookups quick while expanding into deeper research only when a problem needs it.

Less noise, more signal

Automatically sends each query to sources most likely to have fresh, reliable answers.

Works anywhere

Works as an API or MCP server in Claude Code, Cursor, Windsurf, and more.

Works with your favorite tools

Drop-in support for the agents you already use.

For coding agents & developers

Search

Search many sources

Hit documentation, issues, forums, and more in parallel so you see the full picture faster.

Adaptive search depth

Keeps simple lookups quick and only fans out into deeper multi-source research when it actually helps.

Smart source routing

Automatically routes each query to the right sources in parallel, boosting speed and answer quality.

Integrate

Works with your agents and apps

Use the API or MCP server with Claude Code, Cursor, Windsurf, Goose, OpenCode, Codex, and more.

Context-rich responses

Get back not just answers but source links, code snippets, and explanations your agents and apps can use immediately.

Up-to-date information

Search across live documentation, recent releases, and current best practices instead of stale training data.

Build

MCP & API access

Integrate via Model Context Protocol or REST API into your existing agent infrastructure.

Framework agnostic

Use with LangChain, Mastra, Vercel AI SDK, or call directly from your custom agent loop.

Observable search

See exactly what was searched, which sources were scraped, and how results were ranked.

Simple pricing

Community

$0/month

25 free credits

Access to Operator playground

Community & docs support

Developer

$19/month

300 credits

Good for individual developers

Email support

Pro

$299/month

5000 credits

Built for teams and heavy agent usage

Priority support


Common questions

What is Operator?
Operator is an agentic search tool designed specifically for coding agents. It helps Claude Code, Cursor, Windsurf, Goose, and other AI coding assistants find answers when they get stuck on bugs, library issues, or unfamiliar APIs.
How is this different from regular web search?
Operator uses multi-source search, advanced scraping, and difficulty-aware breadth expansion. Instead of returning 10 blue links, it intelligently adapts search depth to the query. Simple lookups are fast, complex debugging problems trigger thorough multi-source research.
What data does the API return?
The API returns structured context including: extracted code snippets, summarized explanations, source URLs with relevance scores, and metadata like publish dates and author info. Everything is formatted for easy consumption by LLMs.
What sources does Operator search?
For code questions: official docs, GitHub issues/discussions, Stack Overflow, and technical blogs. For outage/status questions: status pages, X/Twitter, and Hacker News. For security questions: CVE databases, security advisories, and patch notes.
Which coding agents are supported?
Operator works with any agent that supports MCP (Model Context Protocol) or can make API calls. This includes Claude Code, Cursor, Windsurf, Goose, OpenCode, Codex, GitHub Copilot, and custom agents built with LangChain, Mastra, or Vercel AI SDK.
How does pricing work?
We offer three tiers: Community ($0/month with limited playground access), Developer ($19/month with 300 credits), and Pro ($299/month with 5000 credits and priority support). Harder queries that need more context use more credits.
newsletter layer imagenewsletter layer imagenewsletter layer image

Get started