convez.ai
MCP ServerProdutividadeAvançado
yhooch

Project Delivery Manager

yhooch/project-delivery-manager

Gerenciamento integrado de requisitos, tarefas e bugs com MCP e workflows customizáveis

Ver no GitHub
InicialPluginScore · de 0 a 100

Plataforma SaaS de gerenciamento de entrega de produtos para pequenos times, integrando requisitos, tarefas, bugs, workflows customizáveis e acesso a agentes IA via MCP. Mantém documentação de contexto, boards em tempo real e rastreamento de exceções em um único monorepo TypeScript.

Instalação

$npx convez-ai yhooch/project-delivery-manager

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Equipes de produto coordenando requisitos, tarefas e bugs em um único lugar
  • Automação de workflows customizados com permissões baseadas em papéis
  • Integração de agentes IA (Claude, etc) via MCP para acesso a contexto de projetos
  • Rastreamento de exceções e riscos de entrega em tempo real

Veredito

Solução madura e completa para times de produto que precisam centralizar requisitos, tarefas e bugs com segurança SaaS. O suporte nativo a MCP e integração com agentes IA é um diferencial valioso para automação e contexto aumentado em workflows.

README

A tradução para o português deste README ainda está na fila.

Project Delivery Manager

English | 简体中文

Project Delivery Manager, or PDM, is a SaaS-ready product delivery management platform for small software teams. It brings requirements, intake, tasks, bugs, project context documents, custom workflows, version boards, realtime refresh, and MCP-based AI agent access into one TypeScript monorepo.

Keywords: product delivery management, project management, requirements management, document context library, issue tracking, bug tracking, workflow management, SaaS, MCP, OAuth, Next.js, NestJS, Prisma, PostgreSQL.

What It Solves

Small product teams often keep requirements, decisions, development tasks, test feedback, bug fixes, release scope, and delivery risks in separate tools or chat threads. PDM is designed to keep those objects connected and traceable without turning the workflow into a heavy enterprise process platform.

The core workflow is:

Register/login -> Organization -> Project space -> Version
  -> Requirement / Intake item -> Task / Bug
  -> Workflow action -> Timeline -> Workbench / Board / Exception view

Features

  • SaaS-ready organization and project-space isolation.
  • Username/password authentication with database sessions and HttpOnly cookies.
  • Organization members, project-space members, and role-based permissions.
  • Version management for delivery scope and progress tracking.
  • Requirement documents with Tiptap rich text, Markdown support, image uploads, comments, attachments, and timeline history.
  • Space-level document context library for project knowledge capture, with Markdown/plain text paste, Markdown/Word/HTML import, document list, search, filters, shared folders, detail reading, lightweight editing, re-import, archive/restore/delete, linked resources, tags, comments, attachments, timeline history, Mermaid diagram rendering, and fullscreen Mermaid preview.
  • Intake pool for collecting work sources and splitting accepted items into one or more tasks.
  • Unified work item model for TASK and BUG, with separate user-facing pages.
  • Configurable lightweight workflow engine with states, actions, permissions, action forms, and workflow versions.
  • My workbench, project overview, version board, and exception views.
  • Exception tracking for overdue, blocked, pending confirmation, pending regression, and stale work items.
  • Tags for requirements, intake items, tasks, bugs, and documents.
  • Human-readable object codes such as REQ-12, INTAKE-8, TASK-42, and BUG-17.
  • Realtime local refresh through SSE invalidation events.
  • Chinese and English UI with light, dark, and system themes.
  • MCP server integration for AI agents, protected by OAuth 2.1 + PKCE and scoped Bearer tokens, including document and folder tools.

Screenshots

My workbenchVersion board
My workbench with todos, workflow actions, and recent activityVersion task and bug board with workflow columns
Workflow configurationTask list
Configurable task and bug workflowsTask list with status filters and version metadata
Document context library
Document library with folders, search, import actions, and project context documents
MCP client authorizationsOrganization settings
MCP client authorization list with scoped accessOrganization profile and member management
Account access
Workspace sign-in page

MCP And AI Agent Access

PDM exposes selected product delivery capabilities through Model Context Protocol tools. The MCP surface includes context lookup, object code lookup, workbench, space overview, version board, exceptions, requirements, intake items, tasks, bugs, documents, shared document folders, comments, tags, and timelines.

Current tool names use the pdm.* namespace, for example:

  • pdm.context.get
  • pdm.object.lookup_code
  • pdm.requirement.create
  • pdm.work_item.execute_action
  • pdm.bug.create
  • pdm.document.create_from_markdown
  • pdm.document_folder.create
  • pdm.tag.replace_assignments

MCP requests do not reuse the web session cookie. They require OAuth access tokens scoped for the MCP resource, and business permissions are still enforced by the existing organization, project-space, object, role, and workflow rules.

Tech Stack

AreaTechnology
Monorepopnpm workspaces
WebNext.js, React, TypeScript
APINestJS, TypeScript
DatabasePostgreSQL, Prisma
Shared contractsZod schemas in packages/shared
UITailwind CSS, Radix UI primitives, lucide-react
EditorTiptap plus Markdown mode
File storageMinIO behind the API
RealtimeServer-Sent Events
E2EPlaywright
Unit testsVitest

Repository Layout

apps/
  api/        NestJS API application
  web/        Next.js web application
packages/
  shared/     Zod schemas, DTOs, enums, errors, OpenAPI helpers
prisma/       Prisma schema and migrations
tests/e2e/    Playwright API and UI E2E tests
deploy/       nginx production entrypoint config
docs/         README screenshot assets

Requirements

  • Node.js 22 or newer.
  • Corepack with pnpm 11.1.1.
  • PostgreSQL for the API database.
  • MinIO or another S3-compatible service for attachment flows.

For full local E2E, see tests/e2e/README.md.

Local Development

Enable the pinned package manager:

corepack enable
corepack prepare pnpm@11.1.1 --activate

Install dependencies:

corepack pnpm install

Prepare environment files:

cp .env.example .env

Generate Prisma client and run migrations as needed:

corepack pnpm --filter @project-delivery/api prisma:generate
corepack pnpm --filter @project-delivery/api prisma:migrate

Start both applications:

corepack pnpm dev

Or run them separately:

corepack pnpm dev:api
corepack pnpm dev:web

Verification

Run the standard local gate:

corepack pnpm verify

verify runs lint, typecheck, unit tests, and build.

Run the local E2E gate:

corepack pnpm verify:e2e:local

Run Docker-backed E2E:

corepack pnpm test:e2e:full

Useful targeted commands:

corepack pnpm lint
corepack pnpm typecheck
corepack pnpm test
corepack pnpm build
corepack pnpm test:e2e:list

Deployment

The repository includes a single-machine HTTP deployment path based on Docker Compose:

For HTTP deployment, SESSION_COOKIE_SECURE=false must match the actual origin. When moving to HTTPS, switch it back to secure cookie behavior. Public and private browser entry points are accepted when the request Host matches the browser Origin. OAuth/MCP metadata is derived from the current request Host, and attachment upload/download flows go through the API to the internal MinIO service. Production only needs the Nginx port exposed.

Current Status

The completed and accepted scope includes the core MVP workflow, TAG tagging, SEQ human-readable object codes, RT realtime local refresh, MCP AI agent system access, and the DOC document context library. Remaining release gates are full API/UI E2E against real dependencies, plus target-environment review for RT, MCP, and DOC behavior.

License

This project is licensed under the MIT License.

Ver o original no GitHub

Bom para

Relacionados