Awesome Self-Hosted AI

Open-source AI you can run on your own hardware or server.

308 projects across local LLMs, chat interfaces, agents, MCP servers, RAG, speech and image generation. A weekly job checks every project against the GitHub API, and anything archived, deleted or silent for over a year moves to UNMAINTAINED.md.

308 projects | 307 active in the last 90 days | 10.6M combined stars | 0 unmaintained | refreshed 2026-10-31

See the live health table for stars, last push and license per project.

Contents

• Agents and Automation (27)
• Chat Interfaces (11)
• Coding Assistants (24)
• Document AI and OCR (10)
• Fine-Tuning and Training (10)
• Home and Personal AI (13)
• Image and Video Generation (13)
• LLM Runtimes and Inference (38)
• MCP Servers and Gateways (19)
• Model Gateways and Routers (11)
• Observability and Evals (20)
• RAG and Vector Databases (24)
• Search and Knowledge (13)
• Speech and Audio (15)
• Vision and Multimodal (5)
• Voice Agents (5)
• Rising (25)
• Source-Available (25)

Agents and Automation

• Agent Development Kit - Python toolkit for building and evaluating LLM agents, with multi-model support.
• AgentScope - Multi-agent framework with tracing, memory and tool use for building and deploying agents.
• Agno - Python framework for building multi-agent systems with memory, knowledge, and a runtime for agent apps.
• Botpress - Platform for building and self-hosting conversational AI chatbots and agents with a visual flow builder.
• browser-use - Library that lets LLM agents drive a real browser to complete web tasks.
• CrewAI - Framework for orchestrating role-based teams of AI agents inside your own Python environment.
• Cua - Sandboxes, SDKs and benchmarks for running computer-use agents against desktop environments.
• DeerFlow - Agent harness that coordinates subagents, sandboxed execution, and memory for long-running tasks.
• Forge - Python framework for tool calling and multi-step workflows on self-hosted LLMs.
• Hermes Agent - Autonomous agent that runs server-side, keeps persistent memory and refines its behavior over time.
• Julep - Server platform for stateful agent workflows with memory, tools, branching and long-running tasks.
• Langflow - Visual builder for AI agents and RAG workflows in Python, with each flow exposed as an API.
• LangGraph - Library for building stateful, long-running agent workflows as graphs with checkpointing.
• Letta - Server for stateful agents that keep long-term memory and run as persistent services.
• mem0 - Memory layer for AI assistants that stores and retrieves user facts across sessions.
• Microsandbox - Local microVM runtime that isolates AI agent and untrusted workloads in sandboxed execution.
• Open Interpreter - Natural-language interface that lets an LLM run code and shell commands on your own computer.
• open-multi-agent - Self-hosted TypeScript agent runtime with durable approvals and verifiable run records.
• OpenManus - General-purpose agent framework that runs tools and a browser on your own machine.
• OpenShell - Runtime for autonomous agents that enforces policy-driven execution boundaries, with a CLI.
• Paperclip - Open-source MIT app for coordinating AI agents across projects, tasks and workflows.
• Pydantic AI - Type-safe Python agent framework with model-agnostic tool calling and structured outputs.
• Sim - Visual workflow builder for agents and LLM pipelines that can be self-hosted with Docker.
• Steel Browser - Open-source browser sandbox with an API that lets AI agents run web automation sessions.
• Trigger.dev - TypeScript background job framework for long-running AI tasks with retries and run monitoring.
• UI-TARS Desktop - Desktop app that runs ByteDance's multimodal GUI agent to control a local computer.
• Zep - Memory service that stores and retrieves facts and conversation history for AI agents over time.

Chat Interfaces

• AnythingLLM - Self-hosted app for chatting with documents using local or cloud models and built-in RAG.
• big-AGI - Multi-model chat app with personas and tools that can be self-hosted for power users.
• Chatbox - Cross-platform desktop and web client for chatting with local and remote LLM APIs.
• Cherry Studio - Desktop client for chatting with many LLM providers, with knowledge bases and local document search.
• HuggingChat (self-hosted) - Self-hostable chat web app that connects to OpenAI-compatible model endpoints.
• Jan - Open-source desktop app for running local models and chatting with cloud providers.
• LibreChat - Multi-provider chat UI with plugins, file uploads, and agents that runs in Docker on your server.
• NextChat - Lightweight web chat client for many LLM providers, deployable as a single container.
• Perplexica - AI answering search engine that cites web sources and runs with local or cloud models.
• SillyTavern - Customizable web frontend for chatting and roleplaying with locally hosted or remote LLMs.
• text-generation-webui - Gradio web UI for running local models through llama.cpp, Transformers, and ExLlama backends.

Coding Assistants

• AiderDesk - Desktop app that wraps the Aider coding assistant with a UI, agent mode and MCP support.
• Archon - YAML-defined workflow engine that makes coding agent processes repeatable from planning to validation.
• Claude Code Local - MLX server exposing the Anthropic Messages API so the Claude Code CLI can run against local models on Apple Silicon.
• Cline - VS Code agent that plans and edits code using model API keys you supply.
• Codex CLI - Local terminal coding agent that reads, edits, and runs code inside your repository.
• Continue - IDE extension for autocomplete, chat, and agentic edits against local or hosted models.
• DeepSeek Harness - Plugin-based agent harness for DeepSeek models with a local Web UI.
• DeepSeek-Reasonix - Terminal coding agent built around DeepSeek models, tuned for prefix-cache stability.
• Gemini CLI - Open-source terminal agent for coding and shell tasks, built on Google Gemini models.
• goose - On-machine coding agent that runs shell, file, and extension tools with any supported model.
• Goose - On-machine development agent that edits files and runs tasks, with extensible tool support.
• gptme - Terminal agent with local tools that writes code, runs shell commands and browses the web.
• jcode - Rust coding agent harness built for running many agent sessions at once.
• Kilo Code - Open-source coding agent for VS Code and the terminal, with bring-your-own model keys.
• Mistral Vibe - Lightweight command-line coding agent from Mistral for local development workflows.
• Open Code Review - Code review CLI that combines deterministic analysis passes with LLM agents.
• OpenCode - Terminal-native autonomous coding agent for editing and running code from the command line.
• OpenHands - Platform for AI software agents that edit code, run commands, and browse inside a sandbox.
• Pi - Modular terminal coding agent toolkit with a unified model layer and TUI and web front ends.
• PR-Agent (Qodo) - AI pull request reviewer that runs on GitHub, GitLab, Bitbucket, and Azure DevOps.
• Qwen Code - Terminal coding agent tuned for Qwen models and compatible with other OpenAI-style APIs.
• screenshot-to-code - Turns screenshots, mockups, and Figma designs into frontend code using vision-language models.
• SWE-agent - Software engineering agent that resolves GitHub issues by editing code and running commands in a container.
• T3 Code - Minimal web and desktop GUI for driving coding agents such as Codex and Claude Code.

Document AI and OCR

• Docling - Parses PDF, DOCX, PPTX and HTML into structured Markdown or JSON, with OCR and table recovery.
• GLM-OCR - Open OCR model and code from Z.ai for text, table and formula recognition in document images.
• kordoc - Converts HWP, HWPX, PDF and Office files to Markdown, with a CLI and an MCP server.
• LiteParse - Fast, local, model-free document parser for extracting text and layout from documents.
• Marker - Converts PDFs and other documents to Markdown, with table extraction and optional LLM cleanup.
• MarkItDown (Microsoft) - Command-line and Python tool that converts Office files, PDFs, and images into Markdown.
• OpenContracts - Self-hosted platform for annotating legal documents with semantic search and contract analysis.
• OpenMed - Local-first healthcare app for de-identifying clinical text and extracting medical entities.
• [PaddleOCR](