Real examples of what developers build with ContextWire. Every demo below calls the live API.
The code profile searches MDN, DevDocs, Stack Overflow, and GitHub Discussions. Results are compact — ~150 tokens per result, so 10 results fit in 1.5K tokens of context.
MCP config: Point Claude Code to mcp.kept.live/mcp and it gets all 5 tools automatically.
1. Your claim is reformulated into a search query
2. Multiple engines return evidence from authoritative sources
3. Trinity model evaluates evidence and returns a verdict
4. Sources are cited with URLs for human verification
Verdicts: TRUE / FALSE / MIXED / UNVERIFIABLE
Confidence: 0-100% based on source agreement
The /api/research endpoint runs multiple parallel searches across academic and web sources, then synthesizes a structured report. Depths: quick (3s), standard (8s), deep (15s).
22 built-in profiles let you target exactly the sources you need. No more wading through irrelevant results — every query hits the right engines for the task.