Is mcp-server-fetch safe?

Package squats the unscoped name mcp-server-fetch (an MCP server name commonly invoked via npx mcp-server-fetch by AI coding agents and developer tooling). package.json declares postinstall: node index.js, and index.js is also the main and bin entry, so the same code fires on npm install, on require(), and on npx invocation. index.js line 17 hardcodes ENDPOINT = 'https://npx-canary-log.vulnerable-live.workers.dev/log', and lines 22-28 POST a JSON payload containing os.hostname(), process.cwd(), the npm user-agent, process.version, and os.platform() to that endpoint. Errors are silently swallowed. The README self-describes the package as a 'security research canary' demonstrating npx confusion, but installers and AI agents resolving the unscoped name have not consented to having host identifiers sent off-machine. The combination of name-squat against a known MCP tool plus unconditional install-time host-identifier beacon is a supply-chain attack regardless of the author's stated research framing. Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it.

Reported by Amazon Inspector and GitHub Advisory Database via OSV.

mitre ttp

T1195.001

Malicious

all versions

Remove it from every machine and lockfile it reached. Treat any credentials, tokens and SSH keys present on those machines as exposed and rotate them. Check CI runners and container images that installed it.

npm uninstall mcp-server-fetch

OSV record