OpenMake LLM: Self-Hosted AI Workspace for Local and Open-Weight LLMs

Introduction

OpenMake LLM is an open-source, self-hosted AI workspace and agent runtime designed for local, open-weight, and OpenAI-compatible models. It provides a comprehensive environment to coordinate specialized models, autonomous agents, Model Context Protocol (MCP) tools, deep research, and sandbox execution within a single, self-controlled workspace. Key principles include being local-first, self-hosted, multi-model, agent-centric, and supporting Bring Your Own Key (BYOK).

Why Use It and Key Features

Unlike many self-hosted AI interfaces that focus on interacting with a single model, OpenMake LLM excels at orchestrating multiple models, tools, and agents to accomplish complex tasks. It operates on your own infrastructure, ensuring transparency and traceability of each step.

Key features and capabilities include:

Deep Research: Research topics across multiple search sources, generating detailed reports with citations that can be exported as PDF or DOCX.
Multimodal Interaction: Chat with a local model while a separate vision model reads attached images, or request images, speech output, and transcriptions.
Autonomous Agents: Delegate goals to agents that plan, conduct web research, edit files, execute code in an isolated Docker sandbox, and await approval for risky steps.
Extensible Tooling: Connect external services via MCP, install plugins, skills, or custom agents using Claude-code conventions.
Flexible Model Routing: OpenMake does not require a single model to do everything. It allows you to assign specific models for different capabilities, such as text, vision, image generation, speech, and embeddings, ensuring optimal performance and resource utilization.
Persistent Agent Runtime: Agents maintain task state across rounds, with checkpoints, pause/resume functionality, and recovery after server restarts.
Local Execution Bridge: Run agent work in a local folder using the OpenMake Companion app or the OpenMake Code CLI.

Installation

Getting started with OpenMake LLM is straightforward. The quick start script handles prerequisites and sets up the complete stack.

To install:

curl -fsSL https://raw.githubusercontent.com/openmake/openmake_llm/main/install.sh | bash

This command sets up the necessary tools, configures the environment, starts PostgreSQL and Redis, builds OpenMake, and launches it under PM2. After installation, open the provided URL, log in as an administrator, and connect your preferred model, whether a local vLLM or Ollama server, or any OpenAI-compatible endpoint. For detailed instructions and advanced configurations, refer to the Self-Hosting Guide.

Examples

OpenMake LLM streamlines complex AI workflows by breaking down requests into manageable tasks. A simple question goes directly to your chat model, while more complex requests, such as generating an image or performing a multi-step task, are decomposed and run across your assigned models and tools. Your chat model then synthesizes the final answer from the results.

Consider these scenarios:

Researching a Topic: Provide a research query, and OpenMake will break it into sub-topics, search parallel sources, read and summarize findings, and compile a comprehensive report with citations.
Multimodal Chat: Engage in a conversation with your local LLM, and seamlessly attach images that a dedicated vision model will interpret and incorporate into the discussion.
Agent-Driven Development: Hand an agent a development goal. It will plan the steps, research necessary information online, modify files, execute code in a secure Docker sandbox, and prompt you for approval before making critical changes.