Agent♥︎Age
Catalog

ai.marchward/mcp-server

Official

by marchward · TypeScript

Runtime authority for AI agents: credential mediation, spend cap, approval gates, audit log.

ai.marchward/mcp-server (Model Context Protocol)

This MCP server provides “runtime authority” for AI agents by mediating credentials, enforcing an inference spend cap, applying approval gates, and writing an audit log. It sits in front of an agent’s tool calls and operates as a self-hostable, framework-agnostic open layer for tool calling.

🛠️ Key Features

  • Credential mediation using a Marchward key so agents don’t receive real GitHub/Stripe/database credentials
  • Inference spend cap (hard cap referenced)
  • Approval gates before actions run
  • Audit log with tamper-evident records of decisions

🚀 Use Cases

  • Controlling what AI agents are allowed to do at tool-call runtime
  • Preventing agents from accessing underlying service credentials
  • Tracking and reviewing agent decisions via an audit log

⚡ Developer Benefits

  • Framework-agnostic integration for tool-calling runtimes
  • Agent governance elements aligned with agentic-ai security and safety practices
  • Topics include langchain and langgraph support within the ecosystem context

⚠️ Limitations

  • Description excerpt is partial; specific spend-cap and gate mechanics are not fully detailed in the provided data

Topics

agent-governanceagentic-aiai-agent-securityai-agentsai-governanceai-safetyapache-2-0langchainlanggraphllmmcpopen-coreruntime-authoritytool-callingcredential-mediation

Related servers

More in Security