Releases·Tool Release
FLUJO releases MCP Virtual Computer, a Docker-based VM for AI agents
A developer released mcp-virtual-computer, an open-source Docker-based virtual machine that AI agents can operate through MCP tools while humans watch and control it visually. It integrates with Claude Desktop, FLUJO, Goose, or a browser, exposing a 3D view with toggles for networking and the xfce desktop environment. The tool gives agents a sandboxed full computer to work in, an emerging pattern for agent tooling.
11 engagement1 sourceTue, Sep 1, 2026, 11:14 PM
AI Overview
- GitHub repo flujo-app/mcp-virtual-computer provides the Docker-based VM, announced by its author on Reddit
- Agents interact via MCP tools including `terminal_execute` for running commands and `read_file` for UTF-8 files relative to `/workspace`
- Humans get a 3D view where clicking a green cable toggles the network and a coffee mug icon toggles the xfce desktop environment
- Mouse movements and keystrokes execute visibly in the virtual computer
- Compatible with MCP clients such as Claude Desktop, FLUJO, and Goose, or a browser view
Related
Releases·Tool Release
Nodyra: self-hosted Python workflows with a visual editor, MCP, and Docker workers
A developer launched Nodyra in public beta, a self-hosted workflow automation platform where every node is Python code. Users build workflows visually or let an AI agent connected over MCP create and edit them. The project targets practitioners who want code-level control plus Docker/Kubernetes deployment options.
A developer launched Nodyra in public beta, a self-hosted workflow automation platform where every node is Python code. Users build workflows visually or let an AI agent connected over MCP create and edit them. The project targets practitioners who want code-level control plus Docker/Kubernetes deployment options.
3 engagement1 sourceredditFri, Sep 25, 2026, 11:16 AM
Social·Community
User seeks to run Claude Code in Docker for scheduled tasks without GitHub dependency
A developer asks about running Claude Code in a Docker container to set up recurrent tasks on a remote server, citing incompatibility with their non-GitHub git service and security concerns. The post highlights a gap in Claude Desktop's routines feature, which requires GitHub integration.
A developer asks about running Claude Code in a Docker container to set up recurrent tasks on a remote server, citing incompatibility with their non-GitHub git service and security concerns. The post highlights a gap in Claude Desktop's routines feature, which requires GitHub integration.
18 engagement1 sourceredditThu, Jul 30, 2026, 02:04 PM
Releases·Tool Release
Developer open-sources reddit-mcp-ai, a local MCP server for Reddit
A developer released reddit-mcp-ai, an MIT-licensed Model Context Protocol server that lets AI assistants like Claude Desktop and Cursor search Reddit, explore sentiment, and read saved posts. It is local-first with zero tracking, avoiding the need for commercial credentials or plaintext password storage.
A developer released reddit-mcp-ai, an MIT-licensed Model Context Protocol server that lets AI assistants like Claude Desktop and Cursor search Reddit, explore sentiment, and read saved posts. It is local-first with zero tracking, avoiding the need for commercial credentials or plaintext password storage.
3 engagement1 sourceredditThu, Aug 20, 2026, 01:22 PM
Products·launched
Fact Extract — local RAG and desktop document review tool for Windows
Fact Extract is a free, fully local RAG and document review system for Windows, built mostly with Claude. It offers preprocessing, note taking, PDF export, and a local MCP server that connects to Claude Desktop, OpenWork, Goose, and AnythingLLM. It targets users who want private document search and annotation without cloud services.
Fact Extract is a free, fully local RAG and document review system for Windows, built mostly with Claude. It offers preprocessing, note taking, PDF export, and a local MCP server that connects to Claude Desktop, OpenWork, Goose, and AnythingLLM. It targets users who want private document search and annotation without cloud services.
1 engagement1 sourceredditTue, Sep 1, 2026, 04:49 AM
Products·launched
Text-only LLM controls Linux desktop via vision model describing screenshots
A self-hosted agent that lets a text-only LLM (glm-5.2:cloud) operate a full Linux desktop by routing screenshots to a separate vision model for description. The vision model describes each screenshot, and the text-only model decides the next action. It solves the problem of using non-vision models for GUI automation, targeting developers who want to automate desktop tasks with local or cloud LLMs.
A self-hosted agent that lets a text-only LLM (glm-5.2:cloud) operate a full Linux desktop by routing screenshots to a separate vision model for description. The vision model describes each screenshot, and the text-only model decides the next action. It solves the problem of using non-vision models for GUI automation, targeting developers who want to automate desktop tasks with local or cloud LLMs.
46 engagement1 sourceredditTue, Jul 14, 2026, 04:58 PM