metorial

Metorial is an open-source identity and access control layer that sits between AI agents and external systems, managing authentication, permissions, and observability across more than 1,200 integrations covering SaaS tools, enterprise platforms, and custom MCP servers. Rather than wiring each integration individually, developers connect agents through a single API or connection URL while Metorial handles OAuth flows, API key lifecycle, and service account credentials centrally. It plugs into Claude via the Anthropic SDK (demonstrated with claude-sonnet-4-20250514 in the quick-start examples) and supports MCP server connectivity, making it compatible with Claude Code and other agent frameworks. Built-in RBAC, SAML SSO, scoped credentials per agent or environment, and full audit logs covering which agent acted with whose credentials address the security and compliance concerns that arise when agents interact with production systems. The platform is self-hostable, and a companion CLI, TypeScript and Python SDKs, and an MCP debugging utility called Starbase round out the toolkit. Security teams and platform engineers building multi-agent workflows across shared organizational tooling are the primary beneficiaries.

Lo que la gente pregunta sobre metorial