Open source alternatives to Claude

What Claude does

Claude is Anthropic’s AI assistant for the web, desktop and mobile. It can search the web, create files with code execution, build Artifacts, use connectors and MCP-based desktop extensions, and keep memory and projects. Pro and Max add Research, Claude Code and Claude in Chrome; organizations use Team and Enterprise, and developers the Claude API. The free plan has a session limit that resets every five hours and varies with demand; paid plans add weekly limits.

Why people look for an alternative

• A training choice on consumer plans. Since an August 2025 terms update, Free, Pro and Max users choose whether their chats may be used to improve Claude, and can change it in Privacy Settings. Incognito chats are never used; chats flagged for safety review and thumbs up or down feedback can be.
• Retention follows that choice. Deleted conversations leave back-end storage within 30 days. With model improvement on, chats may be kept de-identified for up to five years; feedback is kept five years, and chats flagged for Usage Policy violations up to two.
• Different rules for business and API use. By default Anthropic does not train on data from Claude for Work, the API or Claude Gov, and API inputs and outputs are deleted within 30 days unless another arrangement or a legal or policy reason applies.
• Data stored in the US. Anthropic states that data is stored in the US, with traffic routed to the US, Europe, Asia and Australia. On the Claude API the US is the only storage region; residency elsewhere is offered through AWS Bedrock, Google Cloud Vertex AI and Microsoft Foundry.
• Limits and model changes. Free has fewer models than Pro, older API models are regularly retired, and plans can change at Anthropic’s discretion.
• No self-hosted version. Claude runs only as Anthropic’s service or through those cloud platforms.

An open source chat app keeps the interface and chat history under your control, but answers are only as private as the model behind it: with a hosted API, prompts still go to that provider. Open-weight models run locally keep prompts on your hardware, but need a capable GPU or plenty of memory and are generally less capable than the top hosted models.

What to compare in an alternative

1. Models. Which local runtimes (Ollama, llama.cpp, a built-in engine) and hosted providers it supports.
2. Where history lives. Browser, desktop or a server database you back up yourself.
3. Tools. The ones you rely on: MCP servers, document chat, code execution, artifact-style previews, web search, projects.
4. Single user or team. A personal desktop app, or a web app with accounts, SSO and roles.
5. Telemetry and license. What the app reports by default, and whether everything is under an OSI-approved license.

Which one to choose

For a shared, self-hosted assistant for a team, LibreChat supports Anthropic alongside OpenAI, Google and local models, with agents, MCP, web search and SSO. AnythingLLM centers on document chat and agents; the MIT-licensed Docker server is its open source edition, and the desktop download has separate terms. For personal use, the desktop clients Jan (which can also run models) and Chatbox connect to hosted APIs and local models. NextChat and Big-AGI are lightweight web clients; Big-AGI can compare and merge answers from several models. Hugging Face’s Chat UI runs HuggingChat, and Khoj is built around your own notes and documents. If prompts must never leave your hardware, TextGen runs local models only, PocketPal AI runs small models on a phone, and Ollama or llama.cpp can serve local models to other clients.