Skip to main content
Glama
kiro-frontend-engineer-mcp

kiro-frontend-engineer-mcp

kiro-frontend-engineer-mcp

A Model Context Protocol (MCP) server implementing a 7-stage agentic frontend workflow:

  1. Design Audit — Web design inspection checklist (accessibility, layout stability, edge cases)

  2. Component Blueprint — AI-driven component generation with full interactive states

  3. Best Practices — React/Next.js guidelines applied before code generation

  4. Browser Validation — Real browser run via kane-cli

  5. E2E Testing — Playwright test execution with pass/fail capture

  6. CI Self-Healing — Auto-fetch GitHub Actions failures and generate fixes

  7. PR Review Loop — Auto-resolve PR review comments

Quick Start — Add to Your Project

Add this to your project's .kiro/settings/mcp.json:

{
  "mcpServers": {
    "frontend-engineer": {
      "command": "node",
      "args": ["<path-to-this-repo>/dist/index.js"]
    }
  }
}

Option 2: npx (after publishing to npm)

{
  "mcpServers": {
    "frontend-engineer": {
      "command": "npx",
      "args": ["-y", "kiro-frontend-engineer-mcp"]
    }
  }
}

Option 3: Clone and build locally

git clone https://github.com/<your-username>/kiro-frontend-engineer-mcp.git
cd kiro-frontend-engineer-mcp
npm install
npm run build

Then point your MCP config to the built output:

{
  "mcpServers": {
    "frontend-engineer": {
      "command": "node",
      "args": ["./kiro-frontend-engineer-mcp/dist/index.js"]
    }
  }
}

Related MCP server: jt-mcp-server

What It Provides

Resources (context the agent reads)

URI

Description

mcp://research/web-design-guidelines

WCAG AA checklist, layout stability, responsive breakpoints

mcp://research/react-best-practices

Server/Client components, TypeScript patterns, performance

Prompts (agent personas)

Name

Description

ui-ux-pro-max

Generates production-ready component blueprints from layout specs

Tools (agent actions)

Name

Description

run_kane_cli

Visual browser validation via kane-cli

execute_playwright_test

Run Playwright E2E tests

github_fetch_ci_logs

Fetch CI failure logs for self-healing

github_address_review_comments

Fetch and resolve PR review comments

Orchestration Flow

The agent should chain calls in this order:

Read web-design-guidelines → Call ui-ux-pro-max prompt → Read react-best-practices
→ Generate code → run_kane_cli → execute_playwright_test
→ (if CI fails) github_fetch_ci_logs → fix → push
→ (if PR comments) github_address_review_comments → resolve → push

Prerequisites

  • Node.js ≥ 18

  • gh CLI (for CI logs and PR comment tools) — gh auth login

  • Playwright installed in your project (for E2E tool)

  • kane-cli (optional, for browser validation tool)

License

MIT

Available Tools

4 tools
execute_playwright_testA

Run local Playwright E2E test suites on generated components. Captures terminal logs, test pass/fail results, and error screenshots.

Inputs: testPath (required), headed.

github_address_review_commentsC

Fetch open PR review comments from GitHub to feed back into agent context for automated resolution.

Inputs: owner (required), repo (required), pullNumber (required).

github_fetch_ci_logsA

Fetch CI build/lint failure logs from GitHub Actions. Used for self-healing: agent reads failures and generates fixes.

Inputs: owner (required), repo (required), runId (required).

run_kane_cliB

Orchestrate a real browser run via kane-cli. Accepts a target URL and optional interaction sequences (click, type, scroll) to visually validate rendered components.

Inputs: url (required), interactions.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool serves a distinct purpose: local testing, PR review comments, CI logs, and visual browser validation. There is no overlap, so an agent can easily distinguish them.

    Naming Consistency3/5

    Three tools follow a verb_noun pattern (execute_, github_fetch_, run_), but 'github_address_review_comments' breaks the pattern—it reads as a noun phrase with 'address' as a verb ambiguous. This inconsistency reduces predictability.

    Tool Count5/5

    With 4 tools, the server is tightly scoped to core frontend engineering feedback loops: local tests, review comments, CI logs, and visual validation. No tool feels redundant or missing.

    Completeness4/5

    The set covers the primary stages of a frontend CI/CD workflow (test, review, CI, visual). A minor gap is the absence of a tool to directly apply generated fixes, but the agent can use existing tools iteratively.

    Related MCP Connectors

    Related MCP Servers