ManagerProtocol
Enterprise governance for AI agents
An MCP server that acts as a decision gatekeeper for AI agents. Before an agent edits a file, calls an API, or modifies data, ManagerProtocol evaluates the action against configurable business rules, returns a risk score, and can block the action or require human approval. Published on npm with 170+ built-in rules across 13 domains.
Demo
See ManagerProtocol in action — evaluating agent actions, returning risk scores, and enforcing governance rules in real time.
Features
Risk Assessment Engine
Evaluates every agent action against configurable business rules before execution. Returns risk scores, identifies violations, and can block dangerous operations automatically.
Human-in-the-Loop Approval
High-risk actions trigger approval workflows. The agent pauses, presents the risk assessment, and waits for explicit human authorization before proceeding.
Session Management
Full lifecycle tracking of agent sessions. Register, monitor, and audit every action an agent takes from start to finish.
Development Activity Logging
Tracks features, bug fixes, refactors, and other development work with structured metadata. Creates an audit trail of what was built and why.
CSS Evaluation
Analyzes new CSS rules for duplicates, specificity issues, missing variables, and suggests improvements before code is committed.
Business Rule Presets
Seven built-in presets (development, standard, strict, etc.) that configure rule sensitivity for different environments and risk tolerances.