OpenAgent is an open-source, self-hostable personal AI assistant written in Go under the Apache-2.0 licence, combining large language models, a retrieval-augmented knowledge base, and autonomous agent loops for individuals and teams who want their own assistant running on their own machine.

What it is

OpenAgent is a next-generation personal AI assistant powered by LLMs, RAG, and agent loops, and it ships as a single binary. The project describes itself as an open-source personal AI assistant that brings together model providers, your own knowledge base, and autonomous agent loops in one self-hostable platform. It is written in Go, licensed under Apache-2.0, and catalogued in the Content & Publishing / Documentation & Knowledge Base category. You can connect any model provider, build a RAG knowledge base from your own documents, and let agents browse the web, run code, and call any MCP-compatible tool on your behalf.

The concrete problem it solves is fragmentation. Assembling a personal assistant normally means subscribing to a hosted chatbot, bolting on a separate retrieval stack for your documents, and wiring up yet another framework for autonomous tool use, each with its own account and data handling. OpenAgent consolidates those layers into one self-hostable platform: it connects 30+ model providers and lets you switch between them per conversation without code changes, keeps your documents in a knowledge base you host, and runs the agent loop locally. It lives in the Go and LLM and agent tooling ecosystem, and it takes the place of a hosted assistant plus a separate RAG pipeline and agent harness.

Key capabilities

• Connects 30+ model providers, including OpenAI, and switches between them per conversation with no code changes required.
• Builds a RAG knowledge base from your own documents.
• Runs agent loops with computer-use, browser-use, and coding agent support, letting agents browse the web and run code.
• Calls any MCP-compatible tool on the user's behalf.
• Ships as a single binary with pre-built releases for Linux, macOS, and Windows on x86_64 and arm64, and Windows runs natively without WSL or Docker.
• Provides Tool Management, Usage Analytics, Activity Monitoring, and Detailed Logs surfaces.
• Offers a Docker image, casbin/openagent, run through docker-compose up.

Who uses it and how

• Individuals who want a personal assistant on their own hardware, started by the installer and reached at http://localhost:14000.
• Teams that self-host a shared knowledge base assistant on internal infrastructure through docker-compose up.
• Developers running coding agent and browser-use tasks against their own repositories and documents.
• Windows users who want a native assistant without setting up WSL or Docker.
• Operators who monitor usage through the Usage Analytics and Activity Monitoring views and audit actions through Detailed Logs.

Getting started

The README's quick start runs an installer that downloads the latest release and starts OpenAgent on port 14000; on macOS, Linux, or WSL that is curl -fsSL https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.sh | bash, on Windows PowerShell it is irm https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.ps1 | iex, and a docker-compose up path is also documented. Building from source needs Go 1.25.0+ for the backend and Node.js 20+ with Yarn 1.x for the frontend.

How it compares

In this registry OpenAgent is catalogued alongside other agent and LLM tooling, sharing topic tags such as langchain, chatgpt, and hermes-agent. The facts provided do not name a set of paid products it directly replaces, so no licence, hosting, or cost comparison is drawn here; its distinguishing facts are the Apache-2.0 licence, the self-hostable single binary, and the per-conversation switching across 30+ model providers.

When to use it — and when not to

OpenAgent fits when you want to run the assistant yourself and keep your knowledge base and agent activity on your own infrastructure, and the single binary makes that a low-effort start. A self-hoster still operates a service on port 14000, and anyone building from source must manage a Go 1.25.0+ toolchain and a Node.js 20+ with Yarn 1.x toolchain. It is a poor fit for someone who wants a fully managed hosted assistant and will not run software on their own machine or server; note also that the README excerpt provided is truncated at the model provider list, so the full feature set is not visible from the facts given.

curl -fsSL https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.sh | bash

irm https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.ps1 | iex

# Backend
go build

# Frontend
cd web && yarn install && yarn start

docker-compose up