ai-devtool-mcp-server

Model Context Protocol (MCP) server for AI DevTool workflow tools/resources/prompts.

What This Repo Owns

• MCP transport endpoints (stdio/SSE).

• Tool registry, schema validation, and policy enforcement.

• MCP resources and prompt catalog for code review and repository chat.

Related MCP server: jt-mcp-server

Bounded Context

• Owns MCP compatibility and tool exposure.

• Does not own core business API lifecycle (api) or UI rendering (web).

System Design Diagram

flowchart LR
    Client[MCP Client] --> Server[MCP Server]
    Server --> Tools[Tool Registry]
    Tools --> API[ai-devtool-api]
    Tools --> Worker[ai-devtool-worker]
    Tools --> Contracts[ai-devtool-contracts]

Tool Invocation Flow

sequenceDiagram
    participant C as MCP Client
    participant M as MCP Server
    participant T as Tool Handler
    participant A as API/Service

    C->>M: tool call
    M->>M: validate request schema
    M->>T: dispatch handler
    T->>A: execute domain action
    A-->>T: result
    T-->>M: typed output
    M-->>C: response + trace id

Planned MVP Tools

• repo_index_status

• repo_search

• repo_read_file

• review_diff

• review_pr_summary

• prompt_catalog_list

Local Development

1. npm install

2. npm run typecheck

3. npm run test

4. npm run lint

5. API-backed stdio server: API_BASE_URL=http://127.0.0.1:3001 npm run start:stdio

6. End-to-end local smoke (boots API + MCP and exercises tools): npm run e2e:local

Run and Deployment Notes

• Required runtime env: API_BASE_URL (defaults to http://localhost:3001 if not set).

• Stdio startup command: npm run start:stdio.

• Expected MCP tools in current phase:

  • repo_index_status

  • repo_search

  • review_start

  • review_status

• For local deployment rehearsal, run npm run e2e:local before publishing to preview/staging.

Quality and Safety Requirements

• All tool contracts validated via shared schemas.

• Allowlist-based tool dispatch.

• Audit logging for every tool invocation.