pardnchiu/Agenvoy

A local AI agent harness in Go that runs 24/7 with auto-built tools, MCP, and Ollama

What it solves

Agenvoy addresses the gap between AI chatbots that merely reply and the need for actual completed work on your own computer. Instead of only giving answers, it can research live data, organize files, run multi-step tasks, and schedule recurring automation—while keeping your memory, schedules, tools, and credentials local and private.

How it works

Agenvoy runs as a single Go binary on your machine (as a daemon). It coordinates models, context, tools, real-time data, task routing, memory, schedules, and execution in one workflow. It uses a sandbox (bubblewrap) to run tools safely and can automatically create, test, and store new tools when none exist. It routes each task to the most suitable model, supports a web dashboard for management, and offers hands-free voice interaction. Agenvoy also acts as an MCP server, letting other agents like Claude Code and Codex share the same sandboxed tools. It connects outward to Telegram and Discord so you can access it remotely without exposing your host.

Who it's for

Agenvoy is for people who want AI to complete work rather than just respond, within an environment they control. It's suited for individuals looking to turn research, file work, and recurring reports into automation, for developers who want a self-hosted agent with local data and sandboxing, for teams that want to share tools across AI agents, and for technical operators needing private remote access via Web, Telegram, or Discord.

Highlights

  • Turns conversation into deliverable work: research real-time data, organize files, and complete multi-step tasks.
  • Fills capability gaps itself: creates, tests, and keeps new tools when none exist, ready for reuse.
  • Shares one tool library across agents: Agenvoy, Claude Code, Codex, and others use the same sandboxed tools.
  • Keeps automation running: one-sentence scheduling for recurring tasks that push results.
  • Makes every step visible and controllable: command output streams to the TUI and web dashboard; sensitive actions require confirmation.
  • Connects freely: route models by task, configure image generation, speech-to-text, text-to-speech, MCP servers, and OAuth.
  • Private remote access: local daemon connects outbound to Telegram and Discord without inbound ports.

Related

  • Project
    java-up-up/nexus-agent

    An enterprise-grade AI agent platform that implements a complete loop from document governance and hybrid RAG retrieval to ReAct-based agentic reasoning and MCP tool integration.

  • Project
    Finsys/hawser

    Hawser is a lightweight Go agent that lets Dockhand manage Docker hosts from anywhere. It supports Standard mode (listens for connections) and Edge mode (initiates outbound WebSocket to Dockhand), making it work across LANs, VPSs, and NAT environments.

  • Project
    spring-ai-community/spring-ai-agent-utils

    A Spring AI library that provides Claude Code-inspired tools for file operations, shell execution, web search, and long-term memory to enable agentic workflows in Java.

  • Project
    NousResearch/Hermes-Bot-Mode

    A desktop plugin for Hermes Agent that turns agent profiles into a roster of named bots with individual personalities, group chat capabilities, and automated routines.