convez.ai
MCP ServerAutomaçãoAvançado
datacore-one

Datacore — Framework para Negócios Automatizados com IA

datacore-one/datacore

IA autônoma que roda suas operações diárias enquanto você estratégia

Ver no GitHub
BomPluginScore · de 0 a 100

Framework open-source que transforma Claude e agentes MCP em operações autônomas de negócio. Oferece memória persistente, gestão de tarefas GTD, base de conhecimento Zettelkasten e execução de agentes noturnos — seu dados permanecem privados na sua máquina.

Instalação

$npx convez-ai datacore-one/datacore

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
4
Forks
8
Acompanhando
0
Saúde
87%
Último commit
há 1 dia
Versão
—
Issues abertas
82
Contribuidores
5

Casos de uso

  • Delegação de tarefas repetitivas para execução noturna (pesquisa, conteúdo, outreach)
  • Gestão de conhecimento pessoal com notas interligadas e busca semântica
  • Captura e organização de tarefas usando metodologia Getting Things Done
  • Construção de agentes IA customizados que entendem contexto e histórico do negócio

Veredito

Infraestrutura sólida para quem quer automatizar operações com IA mantendo controle total dos dados. Requer investimento em arquitetura e Python, mas oferece stack completo: execução agendada, memória persistente e orquestração de agentes. Recomendado para empreendedores técnicos buscando escalar sem terceirizar.

README

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

Datacore

Own Your Intelligence.

An open-source framework for building AI-automated businesses. Datacore gives Claude (and other MCP-compatible agents) the context, structure, and autonomy to run day-to-day operations while you focus on strategy.

License: MIT Python 3.8+ Modules DIPs

Quick Start

Option 1: Let your AI install it

Tell Claude Code (or Cursor, Windsurf, OpenClaw):

"Go to datacore.one and install Datacore."

Option 2: CLI

npx @datacore-one/cli init

Sets up ~/Data, clones modules, and configures the MCP server automatically.

Option 3: MCP server only

npx @datacore-one/mcp init

Then add to .claude/mcp.json or .cursor/mcp.json:

"datacore": {
  "command": "npx",
  "args": ["-y", "@datacore-one/mcp"]
}

Then open Claude Code and try /today or /continue. See GETTING_STARTED.md for a full walkthrough.


What is Datacore?

It starts as an extended mind. It becomes an autonomous business.

Stage 1 — Extended mind        ← start here
  AI that knows your work, remembers your decisions, surfaces what matters.
  Persistent memory via PLUR. GTD task management. Zettelkasten knowledge base.

Stage 2 — Autonomous business  ← where most users end up
  Agents run day-to-day operations: content, research, outreach, coordination.
  Queued during the day. Executed overnight. Reviewed in your morning briefing.

Stage 3 — AI business network  ← the horizon
  Agents from different businesses collaborating and exchanging value.

Your data stays on your drive. You control the agents. You set the direction.

At its core, it provides:

  • Autonomous execution -- Delegate tasks to AI agents overnight; wake up to a quality-evaluated briefing
  • GTD task management -- Capture, organize, and delegate tasks using Getting Things Done methodology with org-mode
  • Knowledge management -- Zettelkasten-style notes, wiki-links, and semantic search across your knowledge base
  • Modular architecture -- Install only what you need; extend with community or custom modules
  • Persistent memory -- Powered by PLUR (preinstalled): corrections, preferences, and decisions survive across sessions

How It Works

You capture ideas and tasks
        |
Datacore organizes, links, and indexes them
        |
AI assistants access your knowledge and context via MCP
        |
Agents execute delegated work overnight
        |
You review results in your morning briefing

Prerequisites


Architecture

~/Data/
|
+-- .datacore/                    # System core
|   +-- agents/                   # AI agent definitions
|   +-- commands/                 # Slash commands (workflows)
|   +-- modules/                  # Installed modules
|   +-- lib/                      # Python utilities
|   +-- specs/                    # System specifications
|   +-- dips/                     # Design proposals
|   +-- registry/                 # Agent, command, source registries
|   +-- state/                    # Runtime state (gitignored)
|   \-- env/                      # Secrets (gitignored)
|
+-- 0-personal/                   # Personal space
|   +-- org/                      # GTD system (org-mode)
|   +-- notes/                    # PKM (Obsidian)
|   +-- code/                     # Personal projects
|   \-- content/                  # Generated content
|
+-- [N]-[name]/                   # Team spaces (separate repos)
|
+-- CLAUDE.md                     # AI context (layered, auto-generated)
+-- install.yaml                  # Installation manifest
\-- sync                          # Multi-repo sync script

Key Concepts

Spaces -- Isolated workspaces for different contexts (personal, teams, organizations). Each space has its own GTD system, knowledge base, and journal. Team spaces are separate git repos.

Agents -- AI agent definitions that handle specific types of work: inbox processing, content writing, data analysis, research orchestration, project management, and more.

Commands -- Slash commands that orchestrate multi-step workflows: /today (morning briefing), /continue (resume work), /tomorrow (end-of-day delegation), /wrap-up (session close).

Modules -- Optional extensions that add domain-specific functionality. Install only what you need.

Layered Context -- Configuration files use a four-layer privacy model (public, org, team, private) so you can contribute improvements upstream without exposing personal data.

Memory -- Persistent memory is handled by PLUR, an open-source engram engine that comes preinstalled. Corrections, preferences, and decisions survive across sessions and are injected automatically — no setup needed.


Modules

Public modules available for community use:

ModuleDescription
gtdGetting Things Done -- task capture, inbox processing, org-mode management
nightshiftAutonomous overnight task execution with multi-persona quality evaluation
researchAutomated research pipelines with knowledge extraction and podcast generation
outboxContent routing out of active workspaces -- archive, delivery, publish
datacortexKnowledge graph -- semantic search, graph statistics, link analysis
crmNetwork intelligence -- track entities, relationships, interaction history
meetingsMeeting lifecycle -- standup generation, preparation, transcription processing
mailEmail integration -- Gmail adapter, classification, processing

See the Module Catalog for installation instructions and the full list of available modules.


Documentation

ResourceDescription
Getting StartedQuick walkthrough for new users
Installation GuideComplete setup instructions
ContributingHow to contribute
Module CatalogAvailable modules and space templates
DIP SpecificationsSystem design documents
Agent RegistryAll registered agents
Command RegistryAll registered commands

Contributing

Datacore uses a fork-and-overlay contribution model. Fork the repo, make improvements to public layer files (.base.md), and submit a PR upstream. Your private configuration stays local and is never shared.

See CONTRIBUTING.md for full guidelines.


License

MIT License -- see LICENSE for details.


Datacore is built by Datacore. The AI system that bootstraps itself into existence.

datacore.one · github.com/datacore-one/datacore

Ver o original no GitHub

Bom para

Relacionados