CVE-2026-61546: AnkiMCP turns off DNS-rebinding protection, so a web page can call its MCP tools, upgrade to 0.21.0
The AnkiMCP add-on, which exposes Anki to AI assistants over MCP, disables FastMCP's DNS-rebinding protection before 0.21.0 and needs no token on its default 127.0.0.1:3141 listener. If you visit a malicious page while Anki is open, the page can rebind to loopback, list the MCP tools, read your collection and create or update notes and decks.
What it is
anki_mcp_server/mcp_server.py sets enable_dns_rebinding_protection=False globally, including for the default local Streamable HTTP listener (CWE-346).
The default config enables HTTP on 127.0.0.1:3141 at the root path, and no bearer token, API key, Host allowlist or Origin allowlist is checked before initialize, tools/list and tools/call.
A malicious web page can use DNS rebinding to reach the listener through the victim's browser, read the local collection and run enabled non-destructive create or update tools.
Destructive tools stay hidden unless enabled_destructive_tools is turned on.
How to check
- Check the AnkiMCP add-on version in Anki (before 0.21.0 is affected).
- Check whether http_enabled is on in the add-on config.
How to fix
- Update the AnkiMCP add-on to 0.21.0 or later.
- Until then, turn off the HTTP listener or close Anki when you aren't using it with an assistant.
- Leave enabled_destructive_tools off.
Sources
- criticalCVE-2026-62863: any web page can drive the MoLing computer-use MCP server in SSE mode, upgrade to 0.5.0
- criticalCVE-2026-93680 and CVE-2026-110186: IBM fixes a 9.8 authentication bypass and unauthenticated code execution in Langflow up to 1.12.8, upgrade to 1.13.0
- criticalCVE-2026-104339 and 30 more: IBM's Langflow 1.12.4 and 1.12.5 bulletins fix unauthenticated RCE, three 9.9 code execution bugs, SSRF and path traversal
- criticalCVE-2026-114612: LiteLLM proxy lets an internal user forge an admin token and run commands