ContextWire
Free search API and MCP server for AI agents: 105 engines, 22 profiles, 94.3% SimpleQA accuracy.
About
ContextWire is a search API built for AI agents, with an MIT-licensed MCP server (keptlive/contextwire-mcp). It aggregates 105 search engines (Google, Bing, DuckDuckGo, Wikipedia, GitHub, arXiv, PubMed, and more) behind five MCP tools: ask (sourced answers), search (105 engines, 22 specialized profiles), extract (clean text from any URL), research (multi-source synthesis), and batch_search (parallel queries).
Setup is a single hosted MCP endpoint (contextwire.dev/mcp) added to Claude Code, Cursor, or any MCP client, with no local server needed. The free tier offers 1,000 queries per month with no credit card, and BYOK support lets you bring your own LLM key. The project claims 94.3% accuracy on the SimpleQA benchmark (project-reported).
The honest picture: this is a thin hosted wrapper over search aggregation, so its durability depends on the operator keeping the endpoint alive and the free tier funded. Benchmark claims are project-reported. For production agents, compare against direct search APIs before depending on a small third party.
What it can do
- Five MCP tools: ask, search, extract, research, batch_search
- 105 aggregated search engines with 22 profiles
- Academic paper search across arXiv and PubMed
- Free tier with 1,000 queries per month
Setup
Add the hosted MCP endpoint URL to the MCP client config; no local install needed.
Prerequisites
An MCP-compatible client such as Claude Code or Cursor.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 114; Trending score 68/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.