AINative PRD Generator MCP Server

AINative PRD Generator MCP Server

Generate production-ready Product Requirement Documents with full AINative platform awareness and persistent memory. Enables AI agents to create, validate, and manage PRDs using ZeroDB for cross-session storage and 18 tools including generation, templates, validation, and semantic search.

Category
访问服务器

README

AINative PRD Generator MCP Server

Generate production-ready Product Requirement Documents with full AINative platform awareness and persistent memory.

An MCP (Model Context Protocol) server that helps AI agents and developers create, validate, and manage PRDs. Unlike generic PRD tools, this server knows every AINative service, API endpoint, SDK, and architectural constraint — so your PRDs reference real platform capabilities from day one.

Why This MCP?

Generic PRD Tools AINative PRD Generator
Generic templates Templates with AINative architecture compliance checklists
No memory — PRDs lost on session close ZeroDB persistence — PRDs saved with version history, searchable across sessions
No platform awareness Knows all 22 AINative products, 1,968 API endpoints, 11+ SDKs
One-shot generation Iterative refinement with section-level regeneration
No validation 15 validation rules + API reference verification

Requirements

An AINative account is required for full functionality. The server uses ZeroDB (AINative's data platform) for:

  • Persistent PRD storage with automatic version tracking
  • Semantic search across all your saved PRDs
  • AI-powered PRD generation via AINative's chat completions API
  • Custom template storage

No account yet? The server auto-provisions a free ZeroDB instance on first run. You'll get a claim URL to take ownership of your data.

Get a permanent account:

npx zerodb-cli init          # Interactive setup
# or sign up at https://ainative.studio

Without credentials, the server still works in template-only mode — template rendering, validation, platform discovery, and scoring all function without an account. Only AI generation and memory/persistence features require authentication.

Quick Start

Option 1: npx (recommended)

npx ainative-prd-mcp

On first run with no credentials, the server:

  1. Provisions a free ZeroDB instance (~800ms)
  2. Saves credentials to .mcp.json and .env
  3. Prints a claim URL — visit it to take permanent ownership

Option 2: With existing API key

ZERODB_API_KEY=ak_your_key npx ainative-prd-mcp

MCP Configuration

Add to your Claude Code, Cursor, or Windsurf MCP config:

{
  "mcpServers": {
    "prd-generator": {
      "command": "npx",
      "args": ["-y", "ainative-prd-mcp"],
      "env": {
        "ZERODB_API_KEY": "ak_your_key",
        "ZERODB_API_URL": "https://api.ainative.studio"
      }
    }
  }
}

No API key? Omit the env block — the server auto-provisions:

{
  "mcpServers": {
    "prd-generator": {
      "command": "npx",
      "args": ["-y", "ainative-prd-mcp"]
    }
  }
}

Tools (23)

Generation (4 tools)

Tool Description
prd_generate Generate a full PRD with AI + AINative platform context. Auto-detects relevant services, saves to ZeroDB.
prd_generate_section Regenerate a single section (e.g., just the Technical Architecture) without touching the rest.
prd_refine Refine an existing PRD based on feedback. Version history tracked automatically.
prd_from_issue Generate a PRD from a GitHub issue number.

Templates (4 tools)

Tool Description
prd_list_templates List built-in and custom templates.
prd_get_template Get a template with placeholder variables.
prd_create_template Create a custom template (persisted in ZeroDB across sessions).
prd_render_template Render a template with variable substitution (no AI, deterministic).

Built-in templates:

  • standard — General-purpose PRD with all standard sections
  • ainative-feature — AINative feature PRD with architecture compliance checklist, TDD test plan, and service mapping
  • agent-capability — Agent/MCP server PRD with tool schemas, memory strategy, and hosting plan

Validation (3 tools)

Tool Description
prd_validate Validate against 15 quality rules + AINative architecture constraints.
prd_score Score completeness 0-100 with letter grade (A-F).
prd_check_api_refs Verify all API endpoint and service references actually exist in the platform.

Validation rules include:

  • Structure checks (title, intro, features, acceptance criteria, timeline)
  • Content checks (problem statement, user stories, test plan, security)
  • AINative-specific checks (correct API paths, ZeroDB usage, no third-party memory services)

Memory — ZeroDB-Powered (4 tools)

Tool Description
prd_save Save a PRD as a persistent plan artifact. Returns an ID for future retrieval.
prd_load Load a saved PRD by ID. Use at session start to resume where you left off.
prd_search Semantic search across all saved PRDs. Find by topic, not just keywords.
prd_history Get version history showing exactly how a PRD evolved over time (unified diffs).

What makes this unique:

PRDs are stored as ZeroDB plan artifacts — a purpose-built storage format that:

  • Survives across sessions — close your editor, come back tomorrow, your PRDs are still there
  • Tracks every change — every prd_refine call generates a diff stored in version history
  • Supports semantic search — find PRDs by meaning ("billing features", "agent deployment") not just title
  • Cross-tool access — PRDs saved here are also accessible via the ZeroMemory MCP tools (zerodb_plan_get)

Platform Discovery (3 tools)

Tool Description
prd_list_services List all 22 AINative products with descriptions, API prefixes, SDKs, and pricing.
prd_get_api_catalog Get detailed API information for a specific service.
prd_suggest_stack Given requirements, suggest which AINative services to use with justifications.

Known AINative services: ZeroDB, ZeroMemory, Agent Cloud, AI Kit, Cody CLI, Chat Completions API, Live Streaming, Multimodal Generation, Embeddings API, Echo Developer Program, OpenCap Stack, ZeroInvoice, ZeroCommerce, ZeroPipeline, Browser Agent, Content Workflow, AX Audit, Community Platform, MCP Hosting, Sequential Thinking, Agent402, QNN API

Skills — GitHub-backed, ZeroDB-cached (5 tools)

Tool Description
skill_list List Agent Skills in the skills repo (pulled live from GitHub).
skill_get Get a skill's full SKILL.md body (optionally with all reference files).
skill_get_reference Get a single reference file for a skill, on demand.
skill_search Find the right skill for a task — ZeroDB semantic search, GitHub keyword fallback.
skill_sync Mirror skills from GitHub into ZeroDB for semantic search + offline use.

Skills are also exposed as MCP prompts: every skill in the repo shows up as a selectable prompt (name = its slug), with an optional input argument for the task to apply it to.

Source of truth is the GitHub repo (SKILLS_REPO, default the8genc/ai-8gent-skills) laid out as skills/<slug>/SKILL.md + skills/<slug>/references/*.md. ZeroDB is a cache and semantic-search layer — author skills in GitHub, then skill_sync to refresh the mirror. skill_list / skill_get work without any credentials; skill_search (semantic) and skill_sync use ZeroDB.

Transports & Hosting

The server speaks MCP over two transports:

  • stdio — default for local use (npx ainative-prd-mcp)
  • Streamable HTTP — auto-selected when $PORT is set (Railway), or forced with MCP_TRANSPORT=http. Serves POST /mcp plus a GET / health check.

Hosted at https://ainative-prd-mcp-production.up.railway.app/mcp:

{
  "mcpServers": {
    "prd-generator": {
      "type": "http",
      "url": "https://ainative-prd-mcp-production.up.railway.app/mcp"
    }
  }
}

Examples

Generate a PRD for a new feature

> Use prd_generate to create a PRD for adding webhook notifications to Agent Cloud

Result: Full PRD generated with:
- Correct API paths (/api/v1/agents/webhooks/*)
- ZeroDB for event storage
- Architecture compliance checklist
- TDD test plan with pytest commands
- Saved to ZeroDB with ID for future sessions

Search past PRDs

> Use prd_search to find PRDs about billing

Result: 3 PRDs found:
- "PRD: Developer Earnings Dashboard" (similarity: 0.89)
- "PRD: Credit System Overhaul" (similarity: 0.82)
- "PRD: Invoice Generation" (similarity: 0.78)

Validate a PRD

> Use prd_validate on this PRD content

Result: Score 73/100 (C)
- Missing: acceptance criteria
- Missing: test plan
- Warning: References "Firebase" — should use ZeroDB instead
- 12/15 rules passed

Authentication

Method Config Notes
API Key (recommended) ZERODB_API_KEY=ak_... Get one: npx zerodb-cli init
Username/Password ZERODB_USERNAME + ZERODB_PASSWORD Auto-refreshes JWT tokens
Auto-provision No config needed Free instance provisioned on first run

Architecture

ainative-prd-mcp/
├── index.js                          # MCP server + auto-provisioning
├── src/
│   ├── client/zerodb-client.js       # ZeroDB API client (auth, plans, memory, chat)
│   ├── tools/
│   │   ├── generation-tools.js       # PRD generation (4 tools)
│   │   ├── template-tools.js         # Template management (4 tools)
│   │   ├── validation-tools.js       # PRD validation (3 tools)
│   │   ├── memory-tools.js           # Persistent storage (4 tools)
│   │   └── platform-tools.js         # Service discovery (3 tools)
│   ├── templates/                    # Built-in Markdown templates
│   │   ├── standard.md
│   │   ├── ainative-feature.md
│   │   └── agent-capability.md
│   └── knowledge/
│       └── platform-manifest.json    # All 22 AINative products/services/APIs
├── .claude/CLAUDE.md                 # Rules for Claude Code agents
├── .cody/CODY.md                     # Rules for Cody/other agents
├── .cody/skills/prd-generator/       # Agent skill definition
└── tests/tools.test.js               # 22 tests

Development

git clone https://github.com/AINative-Studio/ainative-prd-mcp.git
cd ainative-prd-mcp
npm install
npm test              # Run 22 tests
npm run test:coverage # With coverage report

Related

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选