Pracht dev MCP

What pracht dev-mcp does

The pracht dev-mcp MCP server gives coding agents an interface to the resolved application graph of a Pracht project. That graph describes the app's routes, APIs, capabilities, configured agent surface, and build information. Instead of asking an agent to infer application structure from files alone, the server exposes framework-level information and operations through Model Context Protocol.

The server is intended for projects built with Pracht, a Preact framework whose routing, middleware, render modes, APIs, and capabilities are represented in an explicit graph. It is useful during implementation, review, and maintenance when an agent needs to understand how the application is wired.

How it works

Running pracht dev-mcp starts a stdio MCP server. An MCP-compatible coding client can then invoke the server's framework operations while working in the project. The available operations act on the resolved app graph and use Pracht's existing diagnostics, graph diffing, reporting, documentation, and generation workflows.

The graph-oriented design also supports reviewable changes. Agents can inspect the current structure, run checks, compare it with a Git base reference, and work with the committed .pracht/app-graph.json snapshot. This makes route, render-mode, shell, middleware, and API changes visible at the application level rather than only as source-file edits.

Tools and capabilities

The pracht dev-mcp MCP server provides agent access to capabilities described in the project README, including:

• Inspecting resolved routes, APIs, capabilities, agents, and build metadata.
• Running doctor and verify diagnostics, including framework-aware checks and constraint enforcement.
• Planning semantic app-graph differences against a base Git reference.
• Refreshing the app-graph snapshot and producing PR-oriented reports from graph changes, verification results, and bundle budgets.
• Reading Pracht's agent authoring guide through get_docs.
• Scaffolding routes, shells, middleware, and API handlers.

The related command-line workflows include JSON output for inspection and diagnostics, while graph planning can write a refreshed snapshot or produce Markdown for pull requests. Route generation may also create a Playwright smoke test when the project already has an end-to-end setup.

Setup and configuration

Install or use Pracht in the project that an agent will modify, then start the server with:

pracht dev-mcp

Register that stdio process in the MCP-compatible coding client used for the project. The supplied material does not define client-specific registration files, command arguments beyond pracht dev-mcp, or environment variables. Consult the project's MCP documentation for client registration details.

The pracht dev-mcp MCP server operates against the Pracht application's resolved graph, so its most useful results depend on having a Pracht app available for inspection. Scaffolding and verification are intended for an existing project rather than an unrelated codebase.

Limitations and notes

This server is specific to Pracht's application model; the README does not describe it as a general-purpose filesystem, Git, or framework server. Its documented scope is the Pracht graph, diagnostics, reports, documentation, and framework-native scaffolding.

The provided material does not list authentication, remote HTTP transport, configurable environment variables, or a packaged license for the MCP server. It identifies pracht dev-mcp as a stdio server, so deployment and access are tied to the local process registration used by the MCP client.