Claude Code + Barndoor: Governed MCP Access | Barndoor

Govern Claude Code's models, tools, and cost with Barndoor

Enable your team to use Claude Code and build AI agents, governed by Barndoor so every tool, model, and token Claude Code touches is scoped, approved, and logged.

How Claude Code connects through Barndoor

Barndoor gives you one universal endpoint to your MCP servers for Claude Code — connect once, use with all connected MCP servers and approved models.

  1. IT connects your tools at the org level. Authorize Claude Code once in Barndoor.
  2. IT or security define the policy. Which tools, people, and agents; what needs approval; what data is masked; what models are accessed; virtual keys, metering, token budgets.
  3. Point Claude Code at the Barndoor endpoint. One connection governs every tool, model, and agent action.
  4. Give your teams an approved catalog. Of the AI clients and models they’re approved to use.

Why govern Claude Code with Barndoor

Barndoor is the AI Gateway, helping every business use AI and agents securely, reliably, and with control. It governs what Claude Code can access, what it can do with that access, and which models it runs on.

Popular MCP servers to connect with Claude Code

Frequently asked questions

Can I set different budgets for different teams or agents using Claude Code?
Yes. Token and spend budgets can be set per team, per user, or per individual agent, not just at the organization level. This is what lets you track and cap AI cost by who's actually driving it, whether that's a specific developer's agent or a whole department's Claude Code usage.

How do I add a new MCP server for Claude Code after initial setup?
IT or security adds the new MCP server to the approved catalog and sets its policy, and it becomes available to Claude Code and every other connected AI client immediately, no per-client reconfiguration needed.

Why do I need an AI gateway if I already have Claude Code?
Claude Code handles the model interaction, but it doesn't enforce org-wide policy, cost limits, or audit logging on its own. That's what a gateway adds in front of it. Without one, every person or agent using Claude Code is a separate, ungoverned point of access to whatever models and tools they can reach. An AI gateway makes that access centrally scoped, approved, and logged instead.

How does Barndoor protect against unapproved models or tools?
It blocks the request at the gateway before it reaches the model or tool, rather than Claude Code itself deciding to allow or deny it.

Ready to connect Claude Code with Barndoor?