Peta Core

概要

Peta Core functions as a zero-trust Model Context Protocol (MCP) gateway and vault for AI agents, akin to a '1Password for AI agents'. It centralizes security concerns by keeping sensitive credentials encrypted server-side, issuing short-lived agent tokens, enforcing granular access policies with optional human approvals, and logging every tool call. This enables seamless connection of AI clients like ChatGPT, Claude, or custom agents to internal tools, APIs, and data sources without embedding raw secrets into prompts or client configurations, offering a robust solution for secure, observable, and compliant AI agent operations within an organization's infrastructure.

主な機能

• Event persistence for client reconnection and database-backed audit trails with structured logs.
• Zero-trust credential handling with server-side encrypted vault for API keys and secrets.
• 4 GitHub stars
• Three-layer permission model (server, admin, user/client) and human-in-the-loop approvals for high-risk operations.
• Centralized policy enforcement for authentication, authorization, rate limiting, and observability across MCP servers.
• Transparent MCP proxying, acting as an MCP server to clients and an MCP client to downstream servers.

ユースケース

• Connect various AI agents and MCP-compatible clients to internal APIs and data sources without exposing raw secrets.
• Centralize authentication, authorization, rate limiting, and observability for multiple downstream MCP servers.
• Deploy self-hosted or on-premise AI agent infrastructure with encrypted secret storage and audit trails to meet compliance requirements.