Guardrails

---
description: AgentGuards security guardrails for a self-hosted appliance — explains that enforcement is fully automatic via hooks and there is nothing further to call. Load this if asked how guardrails are enforced in this variant, or why no agentguards MCP tools are available.
---

# AgentGuards — self-hosted, hooks-only

This variant bundles no MCP server, so there is no `check_input`/`authorize_action`
to reach for — do not `ToolSearch` for them, they are not installed here.

That is not reduced protection. The three hooks in `hooks/hooks.json` enforce
deterministically, on every request, regardless of anything the model does:

- **`UserPromptSubmit`** screens every prompt before you see it. A block or
  escalate exits non-zero and you never receive the message.
- **`PreToolUse`** authorizes every `Bash` command before it runs.
- **`PostToolUse`** scans content from `WebFetch`, `WebSearch`, and any `Bash`
  command that invokes `curl`/`wget`/`http`/`fetch`/`aria2c` — redacting or
  withholding it before you read it.

Nothing here requires you to call a tool, check a decision, or format a block
message — the hook has already acted by the time you see (or don't see) the
result. If a hook blocks something, its message appears as the tool result or
the reason a prompt was rejected; reply with that message rather than composing
your own.

**Why no MCP server:** Claude Code's plugin `.mcp.json` cannot substitute an
environment variable into its `url` field, so a URL baked in at publish time
cannot be made to point at each installer's own appliance — it would either be
fixed to one instance or, worse, silently fall back to AgentGuards' hosted
service. Rather than ship that, this variant relies entirely on the hooks,
which read `AGENTGUARDS_URL` correctly per installation because they are a
subprocess, not a static config file. If you're on the hosted product instead
of a self-hosted appliance, install `agentguards-claude`, which does bundle the
MCP server.