D

doberman-core

Show HN: Doberman: The AI watchdog that stops Claude from deleting your database

Visit Website

Overview

Doberman is described as adaptive authorization and runtime guardrails for AI coding agents. It is positioned on the tool-execution path rather than inspecting prompts after the model has decided, so a blocked action never runs. It sits between the agent and its tools through a transparent MCP proxy or host hook and turns every action into an explicit, auditable decision. Every tool call receives exactly one verdict before execution: PASS for routine work, AUTH for sensitive actions paused for approval, and BLOCK for dangerous actions that are stopped cold.

It works with Claude Code, Codex, OpenClaw, and any MCP-compatible agent, with Cursor guarded through native hooks as experimental and other MCP clients connecting through the MCP proxy. The project is open source and local first, and it holds two guarantees: it fails closed, meaning uncertainty denies the action, and it is raise-only, meaning it can tighten automatically but never silently loosens. Permanent policy weakening requires explicit, audited human approval gated behind a possession factor.

Additional features include a turn gate that consults the same decision engine before the model starts inferring, a benchmark harness that scores Doberman as a filter over labeled actions, guardrail plugins registered through the doberman.rules entry point, and policy as code via a repo-committed doberman.policy.yaml. The policy file's blocked and sensitive globs fold into every action decision alongside the local role, and policy changes can be reviewed in the same PR as the code they govern.

Pricing

Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.

Visit Website

Prices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.

Use Cases

  • Teams use Doberman to stop dangerous tool calls before they run.
  • Teams use Doberman to pause sensitive actions for human approval.
  • Teams use Doberman to block dangerous actions so they never execute.
  • Teams use Doberman to guard MCP-compatible coding agents through an MCP proxy or host hook.
  • Teams use Doberman to review policy changes in the same PR as the code they govern.
  • Teams use Doberman to forward redacted audit logs to their own pipeline.
  • Teams use Doberman to bound how many decision rows are kept.
  • Teams use Doberman to detect if hook entries are stripped or altered.

Features

Checks AI inputs during runtime

Gates outputs before they proceed

Reviews tool calls before execution

Helps prevent unsafe AI actions

Blocks unintended actions at runtime

Reviews

No reviews yet. Be the first to share your experience!