MCP Integration

How task specs reach your coding agent — and what a coding agent receives when it picks up a task.

Valarchi exposes a remote MCP server at https://valarchi.ai/api/mcp. Connect your coding agent once — then use the MCP Prompt from any task to fetch its full context and start coding.

Setup

Select your coding agent below and follow the instructions to connect it to Valarchi's MCP server. Authentication uses your Valarchi account — you'll be prompted to authorize on first connect.

{
  "mcpServers": {
    "valarchi": {
      "url": "https://valarchi.ai/api/mcp"
    }
  }
}

Next steps:
1. Open Cursor Settings (Cmd/Ctrl + Shift + J)
2. Click on Tools & MCP tab at the side bar
3. Click on New MCP Server button at the bottom of the list
4. Copy the configuration above and paste it into the input field
5. Check MCP server status, Restart Cursor if needed. Is not connecting, use Output tab in bottom bar to see the logs.

Available tools

Once connected, your coding agent has access to eight tools. Start with fetchTaskDetail — it contains everything needed to execute a task. The others exist for deeper context when needed.

fetchTaskDetail

Fetch full details for a specific task — objective, acceptance criteria, wireframe references, DB changes, API endpoints, and test cases. This is the main tool your coding agent should call before starting any task.

Inputs:

productId

milestoneId

taskId (e.g. FEAT-001-API)

fetchTasks

List all tasks for a milestone with IDs, titles, summaries, and dependencies. Use this to get the full task list and pick which task to work on next.

Inputs:

productId

milestoneId

fetchTechnicalPlan

Fetch the complete technical plan including the architecture overview and the full task list summary. Useful for understanding the overall approach before diving into individual tasks.

Inputs:

productId

milestoneId

fetchMilestoneRequirements

Fetch all PRD sections for a milestone — objective, scope, user stories, success criteria, and any custom sections. Use this for product context when the task detail alone isn't enough.

Inputs:

productId

milestoneId

fetchAppStructure

Fetch the confirmed app structure diff — new and modified screens, navigation edges, and layout component changes scoped to this milestone.

Inputs:

productId

milestoneId

fetchWireframe

Fetch the wireframe component code for a specific screen. Task details include wireframe references (screen IDs) — call this for each reference to get the actual component code.

Inputs:

productId

milestoneId

screenId

fetchDesignToken

Fetch design tokens for a product — colors, typography, and spacing extracted from the codebase analysis.

Inputs:

productId

fetchProductVision

Fetch high-level product context from the Master PRD — vision statement, target users, and value proposition.

Inputs:

productId

The Iteration Loop →