Sublime MCP server
{
"mcpServers": {
"sublime": {
"type": "http",
"url": "https://api.sublime.am/mcp"
}
}
}{
"mcpServers": {
"sublime": {
"type": "http",
"url": "https://api.sublime.am/mcp"
}
}
}{
"mcpServers": {
"sublime": {
"type": "http",
"url": "https://api.sublime.am/mcp"
}
}
}{
"servers": {
"sublime": {
"type": "http",
"url": "https://api.sublime.am/mcp"
}
}
}{
"mcpServers": {
"sublime": {
"type": "http",
"url": "https://api.sublime.am/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.0
What it does
A skill that helps AI code agents (Claude Code, Cursor, Copilot, and others) develop procedural agents through a structured six-phase method: discover, redesign, model, build, calibrate, and prove. It integrates with the Sublime MCP server to guide the creation of auditable, measured agents for business processes.
- Discover how a business process actually works by examining concrete examples, then redesign it with explicit trade-offs before automating.
- Build and test procedural agents in preview mode, measuring their behavior against defined correctness criteria before publishing to production.
- Install a shared agent development skill across a team or across multiple code agents on your machine, with version control and project-specific or user-level scoping.
Tools this server exposes
| Tool | Does |
|---|---|
| discover | Discovers how work actually happens in practice from concrete cases. |
| redesign | Defines how the process should work and the decisions about consequences. |
| model | Transforms the process into knowledge, procedures, and decision tables. |
| build | Constructs artifacts in preview and compiles rules before measurement. |
| calibrate | Iteratively measures changes to the agent one at a time. |
| prove | Gathers evidence, reconciles reports, and passes through quality gates before publishing. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.