Jetstack exposes an MCP server over JSON-RPC with support for:
MCP gives Jetstack a structured tool- and resource-oriented integration surface, especially useful for agentic clients and AI-enabled runtime scenarios.
The server negotiates among supported protocol versions and exposes a server-style capability model rather than a simple REST endpoint shape.
Tools expose executable tenant functionality in a structured way. This is especially relevant when another system or agent should perform controlled actions rather than just read data.
Resources expose structured readable content and may also expose templates for parameterized resource retrieval.
Prompts allow prompt discovery and retrieval for agent-facing workflows.
MCP matters operationally because some remote AI providers need public MCP reachability. That makes MCP configuration both an application concern and a deployment concern.
MCP access is not anonymous. It runs under a Jetstack user identity carried by the bearer token used for MCP authentication.
That means MCP access inherits the same core control model as the rest of the platform:
On top of that user-bound identity, MCP tokens may also carry MCP-specific scope restrictions for:
This combination is important. The user identity defines the main permission context, and MCP-specific scopes can narrow the exposed surface even further.
For production MCP integrations, create a dedicated non-human User rather than reusing a personal account.
Recommended setup:
Application access.