Self-Hosted
openclaw-managed-agents
The open alternative to managed agent platforms — any model, any cloud.
★ 441 starsMIT
View guide →A local-first OS for building and running AI agents.
agnt is an open-source local llm project by agnt-gg (agnt-gg/agnt on GitHub). It has 545 GitHub stars and 72 forks, is released under the NOASSERTION license, and was last updated on GitHub on 2 October 2026.
Start at agnt.gg.
Follow the getting-started guide.
Define goals and workflows.
Iterate on agent behavior.
A runtime layer for agents.
Runs on your own machine.
Coordinates multi-agent work.
Repeatable agent pipelines.
Agents that pursue objectives.
Iterate on agent performance.
Download the source and enter the folder.
git clone https://github.com/agnt-gg/agnt.git
cd agntInstall the JavaScript dependencies.
npm installStart the agent OS.
npm startA local-first operating system for building and running AI agents, at agnt-gg/agnt.
It is an agent runtime layer, not a kernel OS.
The code is on GitHub; no license is asserted, so check the repo.
Builders creating agent systems and workflows.
The official site has the getting-started guide.
The open alternative to managed agent platforms — any model, any cloud.
Orchestrate Claude Code, Kiro, and Codex together in isolated tmux sessions.
The fullstack MCP framework to develop MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.
Local-first inbox designed for long-running AI agents, built.
Local-first desktop workspace designed for running.
Local-first personal knowledge layer for AI agents.
The most-starred other Local LLM projects in the LibreAI catalogue — the realistic shortlist if agnt is not the right fit.
The simplest way to run open language models on your own machine.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
LLM inference in C/C++
A high-throughput and memory-efficient inference and serving engine for LLMs
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.
Free and open — grab the source, star the repo, and start creating.