Search Fragments

Turn half-remembered fragments into a named answer: songs, films, books, TV, people, historical events, inventions, sport moments, places, art and culture. Also checks short factual claims. Every answer cites public sources. Free, no account needed.

Hosted MCP Server

npx add-mcp 'https://searchfragments.com/api/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Can it be found? Know before you spend the effort.

A verdict on the half-remembered things a model can’t confidently place.

You half-remember something and can’t place it.

Not for: who directed Jaws · capital of France · well-known facts — normal search is faster there.

For each half-remembered query it returns a verdict: found, with the sources behind it; leads worth following, but not yet confirmed; or not reachable — its search found nothing to continue from. It never fills the gap with a guess.

The gate before the answer

Any question an agent can’t confidently place is a hallucination trap. It carries real signal — a partial memory, an indirect clue, a missing name — but not enough to answer safely from memory. The model recognises the pattern without being able to verify the answer, and a plausible guess becomes a fluent, confidently wrong one.

Search Fragments is the gate between the question and the answer. Before your agent commits — or spends effort digging — it gets a verdict: go, with the evidence behind it; follow these leads, but don’t assert them; or stop, nothing reached — so the agent says “I can’t place this” instead of inventing it. Half-remembered fragments are where it was built; the gate is for any question. Read: Why AI agents hallucinate and confidently invent answers →

Connect to your agent

Available in Claude, Cursor, Raycast and GitHub, or add the URL below to any MCP-compatible agent. When your agent hits a half-remembered fragment it can’t confidently place, it can call Search Fragments instead of guessing. Free. No account needed.

https://searchfragments.com/api/mcp

Manual setup

{
  "mcpServers": {
    "search-fragments": {
      "url": "https://searchfragments.com/api/mcp"
    }
  }
}

URL only.

MCP 2025-11-25 Streamable HTTP · Four tools: gate_probe · resolve_fragment · verify_claim · submit_resolution_feedback

Half-remembered, cross-source queries usually take a dozen tabs and five rephrasings. Search Fragments collapses that into one call.

When an agent hits a half-remembered, cross-source query it can’t pin down, it guesses — confidently, and often wrong. Search Fragments is built for exactly those queries. It returns a verdict first: found with sources, unconfirmed leads to follow, or not reachable by its search. It declines rather than guesses. If your agent has this problem, wire it in and try it — the demo above runs a real resolution right now.

Why Search Fragments exists — one engine, two seekers

  • Found Origin of truth: SF found it and its evidence check passed — a named answer with sources.
  • Leads Partial origin: SF reached real leads it could not confirm. Worth following — not an answer.
  • Not reachable SF’s search reached nothing to continue from. Not a claim that it can’t be found elsewhere.

Built to decline rather than guess.

Tested

We publish both error rates: how often Search Fragments answers wrongly, and how often it wrongly turns away something that could be found. How we test →

Building an agent that hits these queries? Get in touch.