MCP server security scans
An MCP server runs with your agent's access to files, credentials and tools. A bad one can reach your data. Here are scan results for widely used MCP servers. Each page shows the verdict and the kinds of signal we found. The full evidence is in the API report. These are automated signals, and they carry no warranty.
Checking a remote server? Paste its URL into the free MCP server check and see if its tool text is poisoned. No account.
- Filesystem Lets an agent read and write local files.
- GitHub GitHub repository, issue, and pull-request access.
- GitLab GitLab project and repository access.
- Slack Slack workspace messaging access.
- Postgres Read-only PostgreSQL database access.
- Puppeteer Headless-browser automation.
- Brave Search Web search via the Brave Search API.
- Google Maps Places, directions, and geocoding.
- Memory A knowledge-graph memory store.
- Sequential Thinking A step-by-step reasoning scratchpad.
- Everything A test server that uses each MCP feature.
- Playwright Browser automation via Playwright.
- Firecrawl Web scraping and crawling for agents.
- Context7 Up-to-date library documentation for coding agents.
- Tavily Web search and page text for agents.
- Exa Neural web search.
- Notion Notion workspace access.
- Kubernetes Kubernetes cluster operations.
- Browserbase Cloud browser automation.
- Cloudflare Cloudflare account and Workers access.
Scan any server or package: try it · use Lazaretto as an MCP server · API