Skip to main content
Glama
AINative PRD Generator MCP Server

AINative PRD Generator MCP Server

Official

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

Related MCP server: NSTOK AI Software Factory MCP Server

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

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

License

MIT


Powered by ZeroDB + AINative

This package is part of the AINative ecosystem — the AI-native developer platform.

Why ZeroDB?

Feature

ZeroDB

Others

Vector search

Built-in, free embeddings

Separate service (Pinecone, Qdrant)

Agent memory

Cognitive memory with decay + reflection

DIY or Mem0 ($$$)

File storage

S3-compatible, included

Separate S3 bucket

NoSQL tables

Instant, schema-free

MongoDB Atlas, DynamoDB

PostgreSQL

Managed, pgvector pre-installed

Neon, Supabase ($$$)

Serverless functions

DB-event triggered

Firebase/Supabase Edge

Pricing

Free tier, no credit card

Pay-per-query from day 1

Get Started Free

npx zerodb-cli init    # Auto-configures your IDE

Or sign up at ainative.studio — free tier, no credit card required.

View all ZeroDB packages →


Zero-Config Auto-Provisioning

No account needed to start. On first run, this package automatically:

  1. Provisions a free ZeroDB instance (~800ms)

  2. Saves credentials locally (.env, .mcp.json)

  3. Prints a claim link in your terminal

The 72-Hour Window

Your auto-provisioned instance is free for 72 hours. To keep your data permanently, visit the claim URL printed in your terminal.

For AI Agents

If a coding agent (Claude Code, Cursor, Windsurf) installs this package:

  • Auto-provisions without human intervention

  • The claim URL appears in agent output — visit it to take ownership

  • .claude/CLAUDE.md instructs agents to surface the claim link to you

  • Check .env or ~/.ainative/config.json for your project ID if you miss it

Available Tools

23 tools
prd_check_api_refsA

Verify that all API endpoint references in a PRD actually exist in the AINative OpenAPI spec. Catches typos and invalid paths.

Inputs: prd_id, content.

prd_create_templateA

Create a custom PRD template. Stored persistently in ZeroDB so it survives across sessions.

Inputs: name (required), content (required), description (required), sections.

prd_from_issueA

Generate a PRD from a GitHub issue. Fetches the issue title, body, labels, and comments, then expands them into a full PRD with AINative platform context.

Inputs: issue_number (required), repo, template.

prd_generateA

Generate a comprehensive Product Requirements Document (PRD) for an AINative feature, integration, or product. Uses AI with full AINative platform context — knows all services, APIs, SDKs, and architectural constraints. The PRD is automatically saved to ZeroMemory for cross-session recall.

Inputs: description (required), product_name (required), core_features (required), target_audience (required), +5 more.

prd_generate_sectionA

Generate or regenerate a single section of a PRD. Useful for iterative refinement — update just the Technical Architecture or User Stories without regenerating the entire document.

Inputs: section_name (required), prd_id, context, description, +1 more.

prd_get_api_catalogA

Get API endpoint details for a specific AINative service or category. Returns endpoint paths, methods, and descriptions from the OpenAPI catalog.

Inputs: service (required).

prd_get_templateA

Get a PRD template by name. Returns the full Markdown template with placeholder variables.

Inputs: name (required).

prd_historyA

Get version history for a PRD, showing how it evolved over time. Returns unified diffs between versions.

Inputs: prd_id (required).

prd_list_servicesA

List all AINative products and services with descriptions, categories, API prefixes, SDKs, and pricing tiers. Use this before writing a PRD to understand what platform capabilities are available.

Inputs: verbose, category.

prd_list_templatesB

List all available PRD templates (built-in and custom). Each template has a specific structure optimized for different types of products and features.

Inputs: include_custom.

prd_loadA

Load a saved PRD by its artifact ID. Use at the start of a session to resume work on an existing PRD.

Inputs: prd_id (required).

prd_refineA

Refine an existing PRD based on feedback. Provide the PRD ID and feedback, and the AI will update the document while preserving version history.

Inputs: prd_id (required), feedback (required), sections_to_update.

prd_render_templateA

Render a template with variable substitution (no AI, pure placeholder replacement). Fast and deterministic.

Inputs: variables (required), template_name (required).

prd_saveA

Save a PRD to ZeroDB as a persistent plan artifact. Returns an ID you can use to load, update, or search for it in future sessions. Version history is tracked automatically on every update.

Inputs: title (required), content (required), tags.

prd_scoreB

Score a PRD completeness from 0-100 based on section coverage, content quality, and AINative integration depth.

Inputs: prd_id, content.

prd_suggest_stackA

Given a set of requirements, suggest which AINative services, APIs, and SDKs to use. Returns a recommended technology stack with justifications.

Inputs: requirements (required), features, constraints.

prd_validateB

Validate a PRD against quality rules and AINative architectural constraints. Checks for required sections, content completeness, AINative service references, and architecture compliance.

Inputs: prd_id, strict, content.

skill_getA

Get the full content of a skill by slug or name, including its SKILL.md body and (optionally) all reference files. Use this to load a skill before doing the work it describes.

Inputs: skill (required), with_references.

skill_get_referenceA

Get a single reference file for a skill (e.g., "references/testing-patterns.md"). Load reference files on demand when the SKILL.md says to.

Inputs: skill (required), reference (required).

skill_listA

List all Agent Skills available in the skills repo (the8genc/ai-8gent-skills by default). Each entry includes the skill name, slug, description (with trigger phrases), and its reference files. Skills are pulled live from GitHub. Call this first to discover which skills exist before applying one.

Inputs: refresh.

skill_syncA

Mirror skills from the GitHub repo into ZeroDB so they can be semantically searched and served offline. GitHub remains the source of truth — this refreshes the cache. Sync all skills, or one by slug.

Inputs: skill.

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 23 tools

    Disambiguation4/5

    The PRD tools (prd_*) are clearly distinct from each other, covering generation, refinement, validation, search, templates, and more. However, the inclusion of skill_* tools (skill_get, skill_list, etc.) introduces a separate domain that could confuse an agent about the server's primary purpose.

    Naming Consistency5/5

    All tools follow a consistent verb_noun pattern using either 'prd_' or 'skill_' prefix. Within each group, verbs are standard (generate, list, get, save, search) and nouns are descriptive, making the naming predictable and clear.

    Tool Count3/5

    With 23 tools, the server is on the higher end of the acceptable range. While each tool serves a specific purpose, the count feels slightly heavy for a single-purpose server, especially because the skill tools expand the scope beyond PRD generation.

    Completeness4/5

    The PRD tools cover the full lifecycle: creation, generation, refinement, validation, scoring, search, version history, and template management. The only minor gap is the absence of a delete tool. The skill tools are comprehensive but tangential.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      C
      quality
      F
      maintenance
      Enables reasoning agents to turn PRDs into production-ready applications by discovering and reusing existing features, UI components, database models, and templates, then planning, scaffolding, validating, and preparing Git changes for review.
      29
      MIT
    • A
      license
      Not graded
      quality
      A
      maintenance
      Turns GitHub Projects v2 into an agentic project-management platform where AI coding agents can register, claim and track tasks, coordinate via heartbeats, review work, and stay within token budgets. It also enables AI-powered PRD generation, sprint planning, issue triage, roadmap creation, and PRD-to-issue materialization using GitHub-native storage.
      272 npm
      MIT