Quickstart Guide — ContextWire

1

Get Your API Key

Sign up right here to get your free API key instantly. No credit card required.

Response:

JSON Response

{
  "success": true,
  "api_key": "cw_a1b2c3d4e5f6...",
  "message": "Welcome to ContextWire!"
}

2

Make Your First Call

Use the /api/ask endpoint to ask any question. It searches the web, synthesizes sources, and returns a grounded answer.

cURL

curl "https://contextwire.dev/api/ask?q=What+is+the+capital+of+France" \
  -H "Authorization: Bearer YOUR_API_KEY"

Response

{
  "answer": "The capital of France is Paris.",
  "sources": [
    { "title": "France - Wikipedia", "url": "https://en.wikipedia.org/wiki/France" },
    { "title": "Paris | History, Map, Population...", "url": "https://www.britannica.com/place/Paris" }
  ],
  "confidence": 0.97,
  "profile": "web"
}

3

Try Search with Profiles

Profiles tune the search strategy for your use case. Add profile= to any request to optimize results.

🌐 web

💻 code

🎓 academic

📰 news

cURL — Profile Search

curl "https://contextwire.dev/api/ask?q=how+to+parse+JSON+in+Python&profile=code" \
  -H "Authorization: Bearer YOUR_API_KEY"

💡 Each profile uses different source weighting. web is the default and works for general queries. code prioritizes docs and Stack Overflow. academic favors papers and .edu sources. news surfaces recent articles.

4

Extract a Page

Pull clean text, metadata, and structured content from any URL with /api/extract.

cURL

curl "https://contextwire.dev/api/extract?url=https://example.com/article" \
  -H "Authorization: Bearer YOUR_API_KEY"

Response

{
  "title": "Example Article Title",
  "text": "The full extracted article text...",
  "url": "https://example.com/article",
  "word_count": 1247,
  "language": "en"
}

5

Set Up MCP (Claude Code / Cursor)

Connect ContextWire as an MCP server to get grounded search directly inside your AI coding tools.

Add this to your claude_desktop_config.json or .cursor/mcp.json:

MCP Config (JSON)

{
  "mcpServers": {
    "contextwire": {
      "url": "https://mcp.kept.live/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

This gives your AI assistant 5 tools:

ask

Answer any question with sources

search

Multi-source web search

extract

Pull content from any URL

research

Deep multi-step research

factcheck

Verify claims with evidence

6

Install the SDK (Optional)

Use the JavaScript/TypeScript SDK for a cleaner integration.

Terminal

npm install @contextwire/sdk

JavaScript

import { ContextWire } from '@contextwire/sdk';

const cw = new ContextWire('YOUR_API_KEY');

const result = await cw.ask('What is quantum computing?');
console.log(result.answer);
console.log(result.sources);

7

BYOK — Bring Your Own Key

ContextWire uses LLM inference to synthesize answers. By default, this is included free. For higher limits or custom models, pass your own OpenRouter key.

cURL with BYOK

curl "https://contextwire.dev/api/ask?q=explain+CRISPR" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "X-OpenRouter-Key: your_openrouter_key"

🔑 With BYOK, LLM inference costs go to your OpenRouter account. You get unlimited requests with no rate limits on the synthesis step. Search and extraction remain free.

Get Started in 60 Seconds

Ready to build?