Malicious npm packages posing as MCP servers and AI APIs: @shadowdu/memos-mcp, @shadowdqj/image-mcp, mem0-mcp-server and ai-apis
On 16 November 2026 GitHub flagged four more AI-themed npm packages as malware: three named like MCP servers (@shadowdu/memos-mcp, @shadowdqj/image-mcp, and mem0-mcp-server, which uses the name of the Mem0 memory layer) and ai-apis. npm has removed them and replaced each with a 0.0.1-security placeholder. GitHub's guidance: any machine that installed or ran one should be treated as fully compromised.
What it is
@shadowdu/memos-mcp: GHSA-37hh-cmfp-pf7g / MAL-2026-20009, all versions.
@shadowdqj/image-mcp: GHSA-56mh-v6h4-jx2f / MAL-2026-20008, all versions.
mem0-mcp-server (npm): GHSA-9hvg-j8mj-55xr / MAL-2026-20011, all versions. This is not Mem0's own package; check where any Mem0 MCP server you run comes from.
ai-apis: GHSA-4fvr-hmr9-rm3h / MAL-2026-20001, all versions.
The advisories don't describe the payloads. MCP servers are usually started with npx straight from a client config, so a single launch is enough to run the code with your user's access to files, SSH keys, cloud credentials and API keys.
How to check
- Search lockfiles, global installs, the npx cache and MCP client configs (Claude Desktop, Cursor, VS Code, Windsurf) for these four names.
- Look for unknown processes, cron entries, shell profile changes or new SSH keys on machines where they ran.
How to fix
- Remove the packages and their MCP config entries, and clear the npx cache.
- From a clean machine, rotate every secret that was reachable on an affected one: LLM API keys, cloud credentials, GitHub/npm tokens, SSH keys. Rebuilding the machine is the safe option.
- Install MCP servers only from the vendor's documented package name, and pin versions in npx launch lines.
grep -rnE '@shadowdu/memos-mcp|@shadowdqj/image-mcp|"mem0-mcp-server"|"ai-apis"' package-lock.json pnpm-lock.yaml yarn.lock 2>/dev/null
ls ~/.npm/_npx 2>/dev/null | head; grep -rlE 'memos-mcp|image-mcp|mem0-mcp-server|ai-apis' ~/.npm/_npx 2>/dev/nullSources
- criticalMalicious npm package @guicai777/mcp-doctor posed as an MCP diagnostic tool and dropped a hidden Windows executable, with a README telling users to run mshta (MAL-2026-20295)
- criticalCVE-2026-77359: Jupyter MCP Server's /api/connect and /api/stop routes skipped the token and Origin checks, so a website or network caller could repoint it to an attacker's Jupyter backend (fixed 1.0.3)
- criticalMalicious npm packages posing as AI SDKs and MCP servers: character-ai-sdk, char-ai-sdk, @mibbp/agent-observer, mrnima-openai-chatgpt-bot and @fdcasj_dfew/md-doc-mcp
- criticalCVE-2026-76807: Doist's official Todoist MCP HTTP server let any website use DNS rebinding to read, create and delete your Todoist data (fixed 10.3.3)