Agentes de IA de código aberto para delegar tarefas e reunir resultados
MindsHub is a platform for open-source AI agents centered on Cowork, a unified workspace for delegating whole tasks and collecting finished work. Instead of using a chat interface as the final output, it is designed to produce shareable artifacts such as documents, dashboards, apps, reports, and code.
The product combines an agent workspace, a model router, and a credentials vault. MindsHub says this lets you connect your data and tools, choose models from Anthropic, OpenAI, Google, and open-source providers, and run open-source agent harnesses such as Anton and Hermes without managing per-provider API keys. It also supports memory, skills, scheduled tasks, and published artifacts for team handoff and reuse.
Brief a whole task in Cowork and the agent returns finished work instead of a chat transcript. MindsHub frames this around documents, dashboards, reports, apps, and code that you can keep in the workspace or publish by link.
Connect systems through a secure vault so credentials stay scoped per connection. MindsHub lists BigQuery, Postgres, Gmail, Drive, HubSpot, Notion, and Linear as supported systems in the connectors directory.
Use the pre-wired Model Router to choose planning and coding models from a dropdown without per-provider setup. MindsHub says you can switch among frontier models from Anthropic, OpenAI, and Google, or open models such as DeepSeek, Qwen, and Kimi.
Swap between open-source agent harnesses inside the same workspace. MindsHub says Anton and Hermes are available today, with the workspace, data, and artifacts staying in place when you change harnesses.
Turn agent output into reusable artifacts rather than a temporary chat log. MindsHub says results can live in the workspace, be published to a live URL, and be shared with a team by link.
Keep recurring work moving with memory, skills, and scheduled runs. MindsHub says Cowork can remember schemas and preferences, store reusable skills, and run daily briefings, weekly summaries, or recurring audits on a cadence.
Use Cowork when you want an agent to take a larger piece of work and return a finished deliverable instead of a back-and-forth transcript. The product is built around documents, dashboards, reports, apps, and code that can be kept in the workspace or published.
Use the platform for recurring internal work such as daily briefings, weekly summaries, or audit-style tasks. MindsHub says scheduled work lets Cowork run on a cadence, while memory and skills help it reuse prior context and procedures.
Use the connector vault when the work depends on business systems like BigQuery, Postgres, Gmail, Drive, HubSpot, Notion, or Linear. The agent can use scoped credentials at runtime without seeing raw keys.
Use the model router when different steps in one task need different model strengths, such as planning, coding, long context, or faster low-cost drafts. MindsHub says you can switch models from a dropdown and keep the same workspace and history.
Use the standalone agent setup if you want to run an open-source agent with its native harness instead of inside Cowork. MindsHub says Anton and Hermes run in Cowork today, while OpenClaw, OpenClaw + GBrain, NanoClaw, and Hermes are available standalone.
MindsHub Cowork is the unified workspace for open-source agents. You brief a whole task in plain language, connect data and tools, and get back finished work such as documents, dashboards, apps, or code.
The pricing page shows a free Starter plan and a Pro plan at $9.95 per month. MindsHub also says its early-access pricing includes 5M tokens on the paid plan and is billed monthly with cancel anytime terms.
No. MindsHub says the Model Router is pre-wired, so you can start without setting up provider API keys. You can add your own LLM accounts later if you want direct provider control.
Cowork runs on interchangeable open-source agent harnesses. Anton and Hermes are available today, and MindsHub says more are on the way.
MindsHub describes Cowork as a workspace with scratchpads, projects, live artifacts, scheduled tasks, memories, skills, and connectors to systems like BigQuery, Postgres, Gmail, Drive, HubSpot, Notion, and Linear.
Os dados de tráfego são apenas para referência.
browseros.com
BrowserOS é um navegador de IA de código aberto com agentes locais para automatizar tarefas na web usando instruções em linguagem natural. Compatível com macOS, Windows, Linux, extensões do Chrome e integrações MCP.
happycapy.ai
Plataforma de agentes no navegador para executar Claude Code, gerenciar habilidades e delegar tarefas com segurança.
agents.craft.do
Craft Agents é uma aplicação desktop de código aberto para trabalhar com agentes de IA, conectando modelos, APIs, servidores MCP, ficheiros locais e um navegador integrado.
openhands.dev
OpenHands é uma plataforma de código aberto e agnóstica em relação a modelos para agentes de programação na nuvem, que automatiza fluxos de engenharia por meio de interfaces web, CLI, terminal UI e SDK, com implantações locais, na nuvem e empresariais.
trigger.dev
Plataforma de código aberto para jobs em segundo plano, fluxos de IA e agentes em TypeScript
conductor.build
Conductor é um app para Mac que executa agentes do Claude Code, Codex e Cursor em paralelo, em espaços de trabalho isolados, para delegar tarefas, revisar diffs e fazer merges localmente.
