About Realurls
Realurls is a discontinued MCP server for checking whether domains officially belong to software products or companies. Its API and MCP now return “unknown” for every query; the final dataset and audit records remain available read-only.
Key features
- 01Check a product’s official domain
- 02Verify whether a URL is official
- 03Return evidence-based verdicts
- 04Provide user-facing verdict wording
- 05Read-only final dataset and audits
Use cases
- 01Check a download link before sharing it
- 02Assess a pasted login or website URL
- 03Review the final registry and audits
MCP support
The MCP server provides `get_official_url` and `verify_url` tools, plus instructions for hosts. It currently returns “unknown” for all queries.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @realurls/mcpREALURLS_APIAPI base URL (default https://api.realurls.org)Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
2Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| url | string | Yes | URL or bare domain |
Details
From the project README@realurls/mcp
Discontinued on 2026-10-05. Realurls no longer verifies anything: every API and MCP answer is "unknown", and no domain is confirmed as official. Nothing in this repository about a domain should be treated as current. The registry reached 1,000 organizations and every record we audited was correct, but finding the proof that a domain is the real one is slow work per organization, coverage did not grow fast enough to justify it, and no one was relying on the service. What remains is read-only: the final signed dataset (release, CC BY-SA 4.0), the code and rules, and every audit. The service stays up as a notice until 2027-10-05, then is switched off.
MCP server that tells an AI agent which domain officially belongs to which software product or company — so it stops handing users SEO-poisoned lookalike download sites.
Since 0.1.5 this package is a thin stdio bridge to the remote endpoint https://api.realurls.org/mcp: instructions, tool definitions and every answer come from there, so the two ways of installing Realurls can never disagree. Hosts that accept a server URL can use the endpoint directly (Claude Code: claude mcp add --transport http realurls https://api.realurls.org/mcp).
Ownership only. Never a safety judgement. Read TRUST.md for exactly what we do and do not claim.
Tools
| tool | when the agent calls it | returns |
|---|---|---|
get_official_url(name) | before giving any download / login / official-site link | official + verified URLs, or insufficient_evidence / unknown |
verify_url(url) | user pastes a link and asks if it's real | official / not_official (+ the real domain) / insufficient_evidence / unknown |
The server ships instructions telling the host to call get_official_url even when it thinks it already knows the URL.
In our tests that single line is what makes the difference: without it, agents answer well-known tools from memory and never verify.
Every result also carries say_to_user: the sentence the agent should give for that verdict ("I could not confirm the official site", "X is not a known domain of Y; Y's verified site is …"). Only official results contain URLs the agent may hand out.
Install
Claude Desktop — claude_desktop_config.json:
{ "mcpServers": { "realurls": { "command": "npx", "args": ["-y", "@realurls/mcp"] } } }
Claude Code:
claude mcp add realurls -- npx -y @realurls/mcp
Point at a local API during development: REALURLS_API=http://127.0.0.1:8787.
What "official" means
official = the domain has ≥1 anchor (GitHub-verified org, DNS self-attestation, restricted government TLD, …) and ≥2 independent
corroborations, all machine-reproducible. Anything less is reported as insufficient evidence — we would rather say "don't know" than be wrong.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.