Search Fragments — Documentation
Fragment synthesis for agents.
Search Fragments is a remote Model Context Protocol (MCP) server. It resolves under-specified, half-remembered, or fragmentary queries on behalf of AI agents into verified answers — or an honest decline. It is built to decline rather than guess. When it cannot resolve a fragment with confidence, it shortlists honest candidates or declines. Why agents hallucinate on these queries →
What it’s for
Agents frequently hit fragments a normal search can’t resolve cleanly: a half-remembered name, a vague description, a partial reference. Search Fragments takes that fragment, attempts a grounded resolution, and returns one of three honest shapes:
- Resolved — a single confident answer, with the basis for it.
- Shortlist — several ranked candidates when the fragment is genuinely ambiguous, rather than a forced single guess.
- No resolution — an explicit, honest decline when there isn’t enough to go on. No fabricated answer.
The tool: resolve_fragment
- Type: read-only (
readOnlyHint: true) - Input: a text fragment to resolve.
- Output: a resolution result in one of the three shapes above.
- Behaviour: built to prefer an honest shortlist or decline over a forced guess.
Connecting
Remote MCP server using streamable HTTP transport.
- Endpoint:
https://searchfragments.com/api/mcp - Transport: streamable HTTP
- Method: POST (GET/SSE not supported)
- Authentication: none — free, no signup required
- Registry ID:
com.searchfragments/search-fragments
Example client configuration:
{
"mcpServers": {
"search-fragments": {
"type": "remote",
"transport": "streamable-http",
"url": "https://searchfragments.com/api/mcp"
}
}
}A GET request returns a 405 with an explanatory message — the expected healthy response for a browser visit.
Pricing
The service is currently free to use.
Privacy
See /privacy for what data is handled and how. In short: query text is processed to resolve it; tokens are stored only as hashes; query-derived text is sent to Anthropic (Claude) and Tavily.
Support & contact
hello@searchfragments.com — for integration questions or data requests.