Barndoor
Barndoor AI is the control plane for agentic AI, providing secure access and governance for AI agents and Model Context Protocol (MCP) servers. Founded in 2024 by Oren Michels (founder of Mashery), Barndoor enables enterprise IT, security, and developer teams to register agents, govern MCP server access through policy, broker OAuth connections to backend SaaS, and proxy MCP traffic with runtime policy enforcement and full audit trails. The Barndoor Platform REST API manages servers, connections, policies, agents, and MCP / SSE request proxying. Python, TypeScript, and Go SDKs are published on GitHub alongside Rust SDKs (Cerbos, official MCP, MCP OAuth compliance suite) and a Crew AI example. Deployment options include SaaS (trial), private cloud, and on-premises (Enterprise).
APIs
Barndoor Platform API
REST API for the Barndoor Platform. Manage MCP server registrations, OAuth connections from agents to backend SaaS, access-control policies (with rules, restrictions, revisions,...
Barndoor Python SDK
Python SDK for the Barndoor AI Platform. Wraps the Platform REST API, handles Auth0 PKCE login (`loginInteractive()`), discovers governed MCP tools, brokers OAuth connections to...
Barndoor TypeScript SDK
TypeScript SDK for the Barndoor AI Platform. Browser- and Node-friendly client for Auth0 PKCE login, governed MCP tool discovery, OAuth connection initiation, and proxying MCP /...
Barndoor Go SDK
Go SDK for the Barndoor AI Platform. Server-side client for registering agents, managing MCP servers and policies, brokering OAuth connections, and proxying MCP requests from Go...
Official MCP Rust SDK
The official Rust SDK for the Model Context Protocol. Maintained under the Barndoor AI GitHub organization; provides primitives to build MCP clients and servers in Rust.
Cerbos Rust SDK
Rust SDK for Cerbos, the policy-decision-point used by Barndoor for attribute-based access control. Lets Rust services request policy decisions from a Cerbos PDP.
MCP OAuth Compliance Suite
Rust test suite that validates remote MCP servers against the MCP authorization specification - RFC 9728 (Protected Resource Metadata), RFC 8414 (Authorization Server Metadata),...
Barndoor + Crew AI Example
Reference Python demo application showing how to plug Barndoor-governed MCP tools into a Crew AI multi-agent workflow.
Collections
Pricing Plans
Rate Limits
FinOps
Barndoor Finops
FINOPSFeatures
Secure access control and policy enforcement for Model Context Protocol servers.
Continuous governance applied at the moment AI agents act, not just at login.
Precise, scoped access for agents - not broad human-level permissions.
Dynamically surface only policy-compliant MCP tools, optimizing the context window.
Register internal and external agents, group them, and track activity.
Initiate and manage OAuth 2.0 connections from agents to backend SaaS.
Streaming proxy that injects credentials and enforces policy on every MCP and SSE request.
Create, clone, version, validate, and apply Cerbos-based RBAC and ABAC policies.
Complete audit trails for every AI action, applied policy, and outcome.
Stream audit events as gzipped JSON Lines to S3 / GCS / MinIO / SeaweedFS buckets.
Centralized visibility into unauthorized AI apps and agents in the environment.
Connect to existing enterprise IdPs (Keycloak-based) for SSO and identity.
Five dedicated outbound IPs for whitelisting Barndoor traffic at MCP servers.
SaaS, private cloud, and on-premises deployment options for sensitive environments.
Use Cases
Apply access policies and governance to AI agents across the organization.
Centrally register, secure, and manage MCP server deployments for AI agents.
Coordinate multi-agent workflows with security and accountability controls.
Prevent unauthorized AI agent actions and limit data exfiltration.
Surface unauthorized AI apps and agents already running in the environment.
Build agents safely with end-to-end policy enforcement via SDKs.
Govern agents that work across Microsoft 365 (Excel, Outlook, Teams, OneDrive).
Solutions
Centralize AI governance, manage shadow AI, and enforce real-time access controls at scale.
Deploy agents safely without custom security logic, with end-to-end policy across dev, staging, and prod.