AgentGuard

AgentGuard is an MCP server from RLabs for AI code generation quality assurance. It is described as giving Claude, GPT, or Cursor a five-step process to generate production-ready code, and it uses archetypes, or architecture templates, to help keep project structure consistent and quality-focused.

Its documented workflow exposes five capabilities: skeleton for project structure and scaffolding, contracts for type-safe interfaces and schemas, logic for implementation with validation, validate for structural and security checks, and digest for a quality report with trace summary. The page also describes a real example around generating a task management REST API and contrasts outcomes such as inconsistent file structure, partial or missing typing, low test coverage, disorganized imports, and long manual refactoring with the validated, standardized results associated with AgentGuard. It states that the tool provides code verification, protection against hallucinations, granular verification mechanisms, and validation decisions that are explainable and traceable.

AgentGuard is delivered as an MCP server and can be installed with pip as rlabs-agentguard. The page shows a Claude desktop configuration snippet using agentguard-mcp and an archetype argument, and it notes that no API keys or extra configuration are required. It also references a marketplace with 61+ archetypes, 8 built-in archetypes, and 19 MCP tools. Pricing cues on the page include a Get Started Free prompt and monthly installs listed as 878+, while licensing is shown as MIT License. The page title and surrounding copy identify the product as AI quality assurance by RLabs.