# Search Fragments

> Search Fragments is a free MCP server for web search and scraping. Check whether a half-remembered thing can be found before your agent answers. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** API, Web
- **Website:** https://searchfragments.com
- **Repository:** https://github.com/searchfragments/search-fragments-mcp
- **MCP Registry name:** com.searchfragments/search-fragments
- **Categories:** web search and scraping
- **Listing updated:** 2026-11-14

## About

Instead of returning a best guess, this server returns one of three verdicts: found with sources, leads worth following but unconfirmed, or not reachable by its search. The idea is that when a user can only recall a fragment, a model tends to produce a fluent wrong answer, so the agent consults this gate first and says it cannot place something when the verdict is negative.

It is not meant for well-known facts, where an ordinary search is quicker. The authors publish their own tests in both directions. On 50 hard under-documented fragments a baseline agent made three specific fabrications while Search Fragments asserted no answer to any. On 124 graded questions it wrongly asserted an answer in 0 to 2.4 percent of cases, but wrongly declined 40.8 percent of the findable ones. The gate stage let through none of 20 unreachable questions and turned away 4 of 39 findable ones.

It is free, needs no account, and takes just a URL in any client that supports Streamable HTTP. Listings exist for Claude, Cursor, Raycast and GitHub.

## Features

- **Verdict first:** Returns found, leads or not reachable instead of a guess.
- **Gate before answering:** An agent can check reachability before spending effort.
- **Claim verification:** Tests a specific claim against evidence.
- **Published error rates:** Both wrong-assert and wrong-decline rates are given.
- **No account:** Free and keyless, added by URL only.

## Good for

- Identify a film or book from a vague memory
- Stop an agent asserting an unverified name
- Trace a study from a partial description
- Verify a claim before quoting it

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http search-fragments https://searchfragments.com/api/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://searchfragments.com/api/mcp
```

### Cursor (~/.cursor/mcp.json)

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

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "search-fragments": {
      "type": "http",
      "url": "https://searchfragments.com/api/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "search-fragments": {
      "serverUrl": "https://searchfragments.com/api/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "search-fragments": {
      "type": "http",
      "url": "https://searchfragments.com/api/mcp"
    }
  }
}
```

## Tools

- `gate_probe`: The gate before your agent answers. Call this FIRST for any question, before spending effort: one cheap look returns a routing verdict in a few seconds. - QUICK_LOOKUP: one page names the answer, with a quote copied from that page. Check the quote, then answer. - NEEDS_FULL_JOB: worth a full search. Call resolve_fragment with the user's words. - NOT_WORTH_IT: SF's one look found no page naming an 
- `resolve_fragment`: Returns a verdict first, so your agent knows whether to answer, keep looking or say it can't place it: ORIGIN_OF_TRUTH (found, evidence checked, with sources), PARTIAL_ORIGIN (leads to follow, not an answer) or NOT_REACHABLE (SF's search reached nothing; not a claim the thing can't be found). Pieces together the name of a film, TV or radio show, song, book, game, quote, person, place, event or res
- `submit_resolution_feedback`: After the user reacts to a resolve_fragment result, call this with its resolution_id. accepted: it is what they meant. partial: it helped but is not quite it. rejected: it is not what they meant. If they recognised an item in the shortlist or leads, send picked_rank (1 = first item) or picked_url; if nothing fitted, send none_fitted: true. corrected_target and new_clue are optional and are stored 
- `verify_claim`: Checks whether current web sources support a specific factual assertion. Takes a claim as a natural-language statement; returns a verdict with cited evidence and stated limits. Verdict set: - "supported" â€” multiple independent sources directly confirm the specific claimed detail - "partially_supported" â€” sources confirm the entity and domain; specific detail is pointed to but not directly stat

## FAQ

### What are the three verdicts?

Found with sources, leads that are not yet confirmed, and not reachable by its search.

### Is it free?

Yes, with no account required.

### Does a not-reachable verdict mean it does not exist?

No. It only means its own search found nothing to continue from.

### When should I skip it?

For well-known facts, where normal search is faster.

### Is Search Fragments free?

Yes. Search Fragments is free to use.

### How do I add Search Fragments to Claude Code?

Run this command in your terminal: claude mcp add --transport http search-fragments https://searchfragments.com/api/mcp

### What are the best alternatives to Search Fragments?

The closest alternatives to Search Fragments are Scrapling MCP Server, Firecrawl MCP Server, SearXNG Search, NeuralVerge and AI-Powered directory.

## Alternatives

- [Scrapling MCP Server](https://ai-powered.com/mcp/scrapling-mcp-server): Scrapling MCP Server is an open-source MCP server for web search and scraping.
- [Firecrawl MCP Server](https://ai-powered.com/mcp/firecrawl-mcp-server): Firecrawl MCP Server is a freemium MCP server for web search and scraping.
- [SearXNG Search](https://ai-powered.com/mcp/searxng-search): SearXNG Search is an open-source MCP server for web search and scraping.
- [NeuralVerge](https://ai-powered.com/mcp/neuralverge): NeuralVerge is an MCP server for web search and scraping.
- [AI-Powered directory](https://ai-powered.com/mcp/ai-powered-directory): AI-Powered directory is a free MCP server for developer tools.
- [Shopify MCP Server (any store)](https://ai-powered.com/mcp/shopify-mcp-server-any-store): Shopify MCP Server (any store) is a paid MCP server for e-commerce.
- [Chanroll: YouTube Channel Video Scraper (Official API)](https://ai-powered.com/mcp/chanroll-youtube-channel-video-scraper-official-api): Chanroll: YouTube Channel Video Scraper (Official API) is a paid MCP server for design and media.
- [Vidvett: YouTube Video Details Scraper (Official API)](https://ai-powered.com/mcp/vidvett-youtube-video-details-scraper-official-api): Vidvett: YouTube Video Details Scraper (Official API) is a paid MCP server for design and media.

---
Source: https://ai-powered.com/mcp/search-fragments
