The trust layer for MCP servers
Know which MCP servers you can trust.
Thousands of MCP servers, none of them vetted. Vouchity scores every server on the signals that matter — how it's maintained, how widely it's trusted, what it can do to your machine — so you never wire an unvetted tool into your agents again.
320
Servers scored
63
Average Trust Score
269
Carrying a risk flag
11
Categories covered
Synced from the official MCP registry 2 months ago — every score recomputed on each sync.
Why this matters
Most MCP servers ship completely unvetted
Anyone can publish to the registry. Nobody checks who's maintaining it, what it can touch on your machine, or whether it's still alive.
55% ship with no license
176 servers in the registry declare no license at all — no legal permission to use them and nothing Vouchity (or you) can verify about how they're governed before install.
27% grade D or F
85 servers combine stale maintenance, broad host permissions or a missing auth requirement into a score that should give you pause.
269 carry an active risk flag
No-auth remote endpoints, archived repos, deprecated packages, injection-shaped tool text — Vouchity surfaces the exact flag, and keeps watching after you install.
The leaderboard
The most trusted servers right now
Highest Trust Score in the registry today — maintained, widely adopted, and transparent about what they do.
ai.smithery
Browserbasehq MCP Browserbase
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
ai.smithery
Brave
Visit https://brave.com/search/api/ for a free API key. Search the web, local businesses, images,…
ai.foura
MCP
Web scraping for AI agents: fetch any page via proxy or full browser, past anti-bot defenses.
ai.sendinel
Sendinel
AI-controlled email ops for campaigns, contacts, segments, analytics, and sending.
ai.com.mcp
Agent Skills Search Server
Search and discover Agent Skills from the skills.sh registry. Powered by HAPI MCP server.
ai.bowmark
Bowmark
Pre-computed navigation recipes for public websites — skip explore-and-discover.
On the risk watch
Servers you should think twice about
Worst Trust Score first — every card names the exact flag that dragged it down.
ai.buywhere
BuyWhere Product Catalog
Search 1.5M+ products across 20+ platforms. Compare prices, find deals, browse categories.
ai.alpic.test
Test MCP Server
Alpic Test MCP Server - great server!
ac.inference.sh
inference.sh
Run 150+ AI apps — image, video, audio, LLMs, 3D and more. Browse, execute, stream results.
ai.law
Law.AI — Lawyer Search
Verified lawyer and attorney search, discovery, and matching for AI — 991K+ US profiles.
How it works
A score, not a guess
We continuously check every server
Every server in the official MCP registry is re-fetched on a schedule — its repository, its npm package, its license, its activity — so the data never goes stale.
We score five independent signals
Maintenance, adoption, transparency, security and provenance — each weighted and combined into one 0–100 Trust Score, with the evidence for every point shown on the server page.
We alert you when it degrades
Watch the servers you depend on. If a maintainer goes quiet, a repo gets archived or a package is deprecated, you find out before your agent does.
Built for agents, not just humans
Let your agent vet its own tools
An agent shouldn't blindly install whatever MCP server a prompt points it at. Vouchity publishes every Trust Score over a clean JSON API and an llms.txt, so a model can check a server before it ever runs its code.
# Vet a server before installing it
GET https://www.vouchity.com/api/v1/servers/ai-foura-mcp
{
"name": "MCP",
"trust": {
"score": 86,
"grade": "A",
"verdict": "Trusted",
"flags": []
},
"transports": ["streamable-http","stdio"],
"install": "npx @fouradata/mcp",
"remote_endpoint": "https://mcp.foura.ai/mcp"
}Browse by category
Find a trusted server for the job
Code, repos, CI, issue trackers and everything engineers wire into their agents.
SQL and NoSQL stores, warehouses and query engines exposed to agents.
Web search, scraping, crawling and fetch tools that bring the live web to agents.
AWS, GCP, Azure, Kubernetes, Terraform and the infrastructure control plane.
Models, vector stores, RAG, embeddings, NLP and agent-orchestration tooling.
Notes, docs, tasks, calendars and the tools teams run their day on.
Slack, email, Discord, chat and messaging surfaces for agents.
Payments, accounting, markets, SEC filings and financial data connectors.
Headless browsers, computer-use and RPA-style automation servers.
Secrets, scanning, auth, vulnerability and identity tooling.
Documentation, wikis, diagrams, PDFs and knowledge-base retrieval.
FAQ
Good questions
What is Vouchity?＋
Vouchity is a public registry for the Model Context Protocol. Every server published to the official MCP registry gets a transparent, 0–100 Trust Score computed from real GitHub, npm and registry signals — so you can tell which servers are safe to wire into your agents before you install them.
How is the Trust Score calculated?＋
Five weighted signals — maintenance, adoption, transparency, security and provenance — each backed by citable evidence (last commit date, license, download counts, declared auth, and more). A signal that can't be verified from public metadata is excluded from the score entirely, never guessed. See the methodology page for the full breakdown.
Is it free?＋
Yes. Browsing the full registry, the leaderboard, categories and risk watch is free with no account required. Creating a free account lets you build a watchlist and get alerted when a server you depend on changes grade.
Can AI agents use Vouchity?＋
Yes — that's a first-class use case. Vouchity exposes a public Trust API and an llms.txt so a model can look up a server's score, grade and risk flags before it decides whether to install or invoke it.
Stop installing MCP servers blind.
Browse the full Trust Score registry free. Create an account to build a watchlist and get alerted the moment a server you depend on changes grade.