shell-guard

<p align="center"> <h1 align="center">Shell Guard</h1> <p align="center"> <strong>MCP server for safe shell access. Every command validated before execution.</strong> </p> </p>

<p align="center"> <a href="https://agentsid.dev"></a> <a href="https://github.com/stevenkozeniesky02/shell-guard/blob/main/LICENSE"></a> </p>

---

Your AI agent has shell access. It can run ls and cat. It can also run rm -rf / and sudo shutdown now. There's nothing stopping it.

Shell Guard fixes that. Every command is classified and validated against per-agent permission rules before execution. Read commands go through. Destructive commands get blocked. Everything is logged.

Agent runs: ls -la /src
  → Classified as: shell.read.ls
  → AgentsID check: ALLOWED ✓
  → Command executes normally

Agent runs: rm -rf /important-data
  → Classified as: shell.danger.rm
  → AgentsID check: BLOCKED ✗
  → Command never executes
  → Denial logged to audit trail

npx @agentsid/shell-guard

Sign up at agentsid.dev/dashboard (free — 25 agents, 10K events/month).

Register an agent with shell permissions:

npx agentsid init
npx agentsid register-agent --name "my-shell-agent" \
  --permissions "shell.read.*" "shell.git.status" "shell.git.log" "shell.git.diff"

claude mcp add shell-guard \
  -e AGENTSID_PROJECT_KEY=aid_proj_your_key \
  -e AGENTSID_AGENT_TOKEN=aid_tok_your_token \
  -- npx @agentsid/shell-guard

You: "list the files in /src"
Claude: *runs ls /src* → works normally

You: "delete the temp files"
Claude: *runs rm -rf /tmp* → BLOCKED by Shell Guard

You: "push to