Open-source Manus alternatives

OpenClaw is an AI assistant, open source, that you run on your own computer. Instead of living inside a single website, it connects to the messaging channels you already use, including Discord, iMessage, Slack, Teams, Telegram and WhatsApp, with more than twenty others supported according to the project. Native desktop and mobile apps cover macOS, iOS, Android, Windows and Linux.

A single Gateway process runs the assistant, either as a personal setup on one laptop or as a deployment shared across a team, with configuration being the main difference between the two. State, memory and credentials stay on your hardware, and the model or agent harness is a swappable plugin, so you can use hosted models such as Claude or Codex, or local models, without changing the rest of the setup.

OpenClaw is stewarded by an independent nonprofit foundation, and the project says it offers no paid plan, no hosted service and no token. By default it only contacts its servers for a daily version check, and anonymous statistics are opt-in. GitHub lists the license as 'Other', so check the repository for the exact terms. It suits people who want a private assistant they control themselves.

AutoGPT is an open-source platform for building and running AI agents that carry out complete workflows. You can describe a desired outcome in plain English, or lay out each step yourself in a visual builder. Agents can then run whenever you ask, on a schedule or from a trigger. The project is written mainly in Python.

The platform is organized around four areas: AutoPilot, which turns a conversation into a working agent; Agents, which shows runs, costs and actions needing attention; a Marketplace of ready-made agents you can add to your library and customize; and Build, a block-based editor where you connect, branch and inspect steps for precise control.

A hosted platform is run by the vendor, who manages infrastructure, model access, credentials and updates, and it is a paid service. The code can also be self-hosted, and the README links a self-hosting guide. GitHub lists the license as 'Other', so review the repository terms before reuse. It suits teams automating repeatable tasks with agents.

OpenHands is an open-source project for AI-driven software development. Its current README centers on Agent Canvas, a self-hosted developer control center for starting conversations with coding agents and automating everyday engineering tasks. It is written mainly in TypeScript and released under the MIT license.

By default Agent Canvas uses the open-source OpenHands agent, but it can also drive third-party agents such as Claude Code, Codex and Gemini, or any ACP-compatible agent. It runs locally by default and can connect to several agent backends, including Docker containers, virtual machines and company infrastructure, so agents can be moved between local, remote and cloud setups without interrupting your work.

Automations can be triggered on a timetable or by incoming webhook events and connect with tools such as Slack, GitHub, Linear and Notion, for example publishing reports to Slack or breaking GitHub issues into tasks. Agents can optionally run on OpenHands Cloud or OpenHands Enterprise. It suits engineering teams that want to coordinate several coding agents from one self-managed place.

goose is an AI agent for general tasks that runs locally on your own machine. It is not limited to coding: the project suggests using it for research, writing, automation, data analysis and similar tasks. It is built in Rust, released under the Apache-2.0 license, and available as a desktop application on Windows, Linux and macOS, a full CLI for terminal workflows, and an API for embedding it elsewhere.

goose works with more than a dozen model providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure and Bedrock, and can use existing Claude, ChatGPT or Gemini subscriptions through ACP as an alternative to API keys. It connects to many extensions through the Model Context Protocol open standard, and custom distributions let teams ship their own goose build with preconfigured providers, extensions and branding.

The project is part of the Agentic AI Foundation at the Linux Foundation, and its documentation covers a quickstart, installation, tutorials and governance. It suits developers and knowledge workers who want a configurable local agent that is not tied to a single model vendor.

AionUi is an open-source app built around the idea of cowork rather than simple chat: AI agents share your computer with you, opening files, writing code, browsing the web and automating chores, while you can see everything the agent does and stay in control. It runs on multiple platforms.

A built-in agent works with zero setup, and you can bring any API key. For heavier use, AionUi can run many external CLI agents in one interface, including Claude Code, Codex, Qwen Code, Hermes Agent and Cursor Agent, with multi-agent teams and customizable assistants. Remote access is possible through a WebUI and through Telegram, Lark, DingTalk and WeChat, and cron-based scheduling lets automations run unattended around the clock. The project says this contrasts with traditional AI chat clients, which can rarely act on files or execute multi-step tasks.

AionUi is written in TypeScript and licensed under Apache-2.0. It is free and open source, and it fits developers and office users who want agents that act on local files and tools rather than only answering questions.

Suna is the repository of Kortix, which describes itself as an open-source AI Management System. Instead of a single chatbot, it organizes AI agents that carry out tasks and deliver finished work such as reports, code and replies. Agents, the skills they share, company memory and connectors to other services are all kept as files in a git repository, so configuration is versioned and can be reviewed like code.

Each agent session runs in an isolated cloud sandbox on its own branch, and the result lands as a change request that a human approves before it is merged. The platform works with any model provider using your own API keys, rather than being tied to a single model vendor.

The project positions itself as an open alternative to closed, hosted agent products such as ChatGPT, offering source access and the option to run on your own cloud or on-premise network. The repository lists its license as 'Other', so check the license file before commercial use. A managed cloud edition is also available.

LobsterAI is a desktop AI agent built by NetEase Youdao that can work in your real computing environment, including local files, shell commands, browser workflows, office documents, spreadsheets, slides, scheduled jobs and project workspaces. It is aimed at office-style tasks such as data analysis, slide and document creation, video work and web research.

The app is organized around Cowork, its product and session layer, which keeps local persistence, permissions, interface state, artifacts, agents, memory and instant-messaging bindings. OpenClaw is the runtime and gateway beneath it that actually executes agent tasks. Users can send commands from their phone through WeChat, Feishu, DingTalk or Telegram, and MCP is among its repository topics.

LobsterAI is built with Electron, React and TypeScript, and it is licensed under MIT. It runs as a desktop application on your own computer rather than as a hosted service. It suits users who want an agent with real local access and mobile messaging control.

Poco, or Poco Claw, is an open-source AI agent workspace powered by a Claude Code-based agent. It is presented as a more polished and safer alternative to OpenClaw, with a web interface, built-in messaging support and a sandboxed runtime that keeps tasks isolated from the host machine.

Teams work in long-lived servers with channels, direct messages and reply threads, can mention persistent agents, turn messages into trackable tasks and publish outputs to shared files. Every task runs in an isolated container, with optional mounting of local directories when self-hosting. Other features include plan mode, conversation queueing, project settings for model and Git repository, an artifacts view for HTML, PDF, Markdown and diagrams, a playback view for commands and browser sessions, MCP and skills import, a built-in browser and scheduled background runs.

The project is written in Python, licensed under MIT, and has documentation hosted separately. It can be self-hosted, and the README notes that local directory mounting is available only in that mode. It suits developers and small teams who want to delegate coding and research work to a shared, observable agent.

Lemon AI is an open-source general AI agent designed to run on your own hardware. It is presented as a local alternative to agent platforms such as Manus and Genspark AI, and it includes a code-interpreter virtual machine sandbox so code can be written and executed without touching your own files or operating system. It targets users who want deep research, web browsing, coding and data analysis with privacy.

The agent supports planning, action, reflection and memory using local models such as DeepSeek, Kimi, Qwen, Llama, Gemma and GPT-OSS through Ollama or vLLM. For stronger results you can also configure APIs from cloud models including Claude, GPT, Gemini and Grok. Supported tasks include research reports, code generation, data analysis, content creation and document processing. An experience repository lets the agent learn and be customized for an organization.

The sandbox runs inside Docker, and a quick-deployment guide, documentation and a desktop build are available. The codebase is JavaScript running on Node.js with Vue 3. The repository metadata lists the license as Other, so review the license file before commercial use.

Open-Agent is an open-source agentic AI project presented as an alternative to the Claude Agent SDK, ChatGPT Agents and Manus. The idea is that agentic systems can do meaningful work by operating computers, browsers and phones like a person, and that opening them up lets others modify and extend them. You chat with leading models while a multi-agent system works on your tasks, and you can use the project as the foundation for your own dedicated agent.

The key features are a customizable agent setup that integrates OpenAI, Claude, Gemini and open-source models, spec and context engineering that helps agents plan, score and surface options while you keep the final decision, and multi-agent collaboration in which several frontier models work together on a task. The project is open source and free to modify.

Deployment uses Docker Compose: you copy the example configuration and compose file into a folder, add your API keys to the configuration and start the stack. The code is TypeScript and licensed under Apache-2.0. The project welcomes contributions and has a Discord community, and the website open-agent.io provides more details.