agenvoy/Agenvoy

Self-hosted AI agent harness in a single Go binary — writes, sandbox-tests and repairs its own tools, and lets Claude Code, Codex and any MCP client build and share them.

What it solves

Agenvoy is a self-hosted AI agent harness designed to overcome the limitation of static toolsets. It allows agents to build, test, and reuse their own tools in a sandboxed environment, ensuring that when a task requires a capability the agent doesn't have, it can create the necessary tool instead of failing.

How it works

It operates as a single Go binary that functions as both an agent application and an MCP (Model Context Protocol) server. This allows it to maintain a shared tool library that can be accessed by other AI agents like Claude Code or Codex. The system includes a dispatcher for skill-aware model routing, a sandboxed execution environment for security, and integrations for scheduling recurring tasks and searching local files.

Who it’s for

Developers and technical operators who need a self-hosted agent with security guardrails, teams wanting to share reusable tools across different AI agents, and users requiring integrated automation, file search, and scheduled reporting.

Highlights

  • Auto tool generation: Automatically builds and saves missing tools to expand its own capabilities.
  • Shared Tool Library: Acts as an MCP server to share its sandboxed tools with other agents.
  • Self-scheduling: Converts natural language requests into automated cron jobs.
  • Local File Integration: Searches and answers questions based on local files and knowledge notes.
  • Multi-channel Support: Routes approval prompts and feedback across CLI, Web, Telegram, and Discord.
  • Secure Sandbox: Implements file boundaries and confirmation prompts for sensitive path access.

Related

  • Project
    bionic-gpt/bionic-gpt

    An open-source, self-hosted agentic harness for internal AI teams to build sovereign AI workflows with sandboxed code execution and model independence.

  • Project
    ruvnet/metaharness

    A factory for minting custom, repo-aware AI agent harnesses with their own CLI, skills, and governance, compatible with multiple agent hosts.

  • Project
    uvwt/agentdock

    A secure tool runtime for AI agents that enables them to operate files, commands, and deployments across multiple local and remote machines via the Model Context Protocol.

  • Project
    Mng-dev-ai/agentrove

    A self-hosted AI coding workspace that orchestrates multiple AI agents from one interface, providing sandboxed environments and multi-agent workflows.

  • Project
    EverMind-AI/Raven

    Raven is an open-source, self-improving agent harness that integrates long-term memory, proactivity, and local tracing for long-running AI agent workflows.