Build an MCP endpoint that fits your workflow.
Create a workspace, shape the tools and context you want to share, then connect an MCP client to your server.
WorkspaceBuild server capabilities
MCP endpointPublish over Streamable HTTP
MCP clientDiscover and call capabilities
Choose a guide
Follow a setup path or look up one part of your server.
Getting startedStart MCPKite locally with Docker Compose.
Servers and clientsCreate a server and configure an MCP client.
Python toolsWrite, save, and safely publish a tool.
PromptsBuild reusable templates with typed arguments.
ResourcesPublish text, URI templates, and private files.
Workspace settingsManage members, tokens, variables, and secrets.
Deployment and configurationPrepare storage, authentication, and runtime limits.
TroubleshootingFind the right endpoint, setting, or log.
Know the boundary
MCPKite keeps management behind workspace authentication and runs saved Python tools in short-lived sandbox containers. For a hosted deployment, follow the operator guide before exposing the service.
Review the production checklist