Security review

Is context-mode MCP safe to give an agent?

destructive capableneeds cautiondestructive capablehealthy

A factual risk summary built from context-mode’s real tool surface, execution model, and verification history — not a vibe. Trust score 46/100.

01What it can do

Has tools that can delete or irreversibly change data.

11 tools observeddestructive presentwrite present

02Execution model

Runs locally over stdio — the server process executes on your machine with your user's privileges. Vet the source and package before granting access.

Installs via npx (`npx -y context-mode`) — it pulls and executes third-party code; pin versions and review the source.

03Permissions & auth

Knowledge & memory · writeVector & semantic search · writeVersion control (git) · writeScrape a website · writeExecute code · writeQuery a database · writeRead & write files · writeMaps & location · destructive

The write/destructive access this server can exercise, inferred from its verified tools.

04Verification

handshakepassed — tool surface is real

runshandshake:failed · metadata:passed · handshake:failed · metadata:passed · handshake:failed · metadata:passed · tools_list:passed · handshake:passed · metadata:passed · handshake:failed

last_checked2026-08-16 22:30Z

sourcesGitHub repo search [p4]

Reduce the risk

Worried about handing an agent raw access? See governed agents in action — Apex gives your AI paced, capped, fully-logged hands with approval queues before anything runs.

Explore Apex →

See also: full server page · setup · alternatives