What these docs cover
MCP Boundary wraps a local command-based MCP server, discovers tools, shows the tool surface in a local dashboard, and checks concrete tool calls before they reach the downstream server.
Documentation navigation
MCP Boundary docs
MCP Boundary docs
Product documentation for installing MCP Boundary, wrapping local command-based MCP servers, reading the dashboard, and understanding current limits.
Product docs
Get started
- Getting started guideDownload the v0.1.0 package, run the local quickstart, and open the dashboard.
- Add an MCP serverUse a local command such as npx, uvx, python, node, or a local binary as the downstream server.
- CompatibilityUnderstand why local stdio servers work best today and where remote/login-heavy servers need separate support.
Operate the boundary
- Dashboard guideRead Setup, Tools, Activity, and Policy without treating the dashboard as execution authority.
- Policy examplesStart with simple rules: hide tools, allow narrow calls, block risky effects, and inspect activity.
- TroubleshootingCommon startup, discovery, stdout, login, and missing-tool issues.
What stays local
Server configuration, discovered tools, policy files, activity logs, and downstream outcomes stay on the local machine. Public website analytics may count aggregate release downloads only.
Read before use
MCP Boundary is a local developer/operator tool. It is not a DLP system, not a prompt-injection detector, and not a full enterprise security gateway.