AnythingLLM

Mintplex Labs Inc

About this app

AnythingLLM is the open-source AI agent for your phone. It doesn't just chat, it gets things done: researches the web, plans your day, turns conversations into documents, and runs tasks for you on a schedule.

AI you own, not AI you rent. Use any model you like, from the most powerful cloud AI to private models that run entirely on your phone, offline, with no internet at all. Switch anytime. No lock-in, and the code is open for anyone to see.

AN AGENT THAT GETS THINGS DONE
• Researches the web, reads the pages, and answers with sources
• Checks your calendar, adds events, and drafts emails and texts for you to send
• Turns any chat into a Word document, PDF, or slideshow
• Runs tasks on a schedule, like a morning news digest or a weekly check-in, and notifies you when they're done

WORKS EVERYWHERE ON YOUR PHONE
• Highlight text in any app to rewrite, summarize, or explain it instantly
• Share photos, documents, and links from any app straight to your agent
• Get notified when a reply is ready, even with your phone locked

KNOWS WHAT MATTERS TO YOU
• Remembers what you tell it, across every chat or just one workspace
• Understands your PDFs, Word docs, spreadsheets, and slides
• Attach a photo and ask about it
• Talk instead of type with voice input

ANY MODEL. YOUR RULES.
• Connect popular cloud AI services or any compatible provider
• Use AI running on your own computer or AnythingLLM Desktop
• Go fully offline with free open models from Hugging Face, and see which ones fit your phone before you download
• Your chats, files, and memories are stored on your phone, not our servers. You choose which AI sees them.

OPEN SOURCE
AnythingLLM is built in the open by the team behind AnythingLLM Desktop. Read the code, suggest features, or contribute on GitHub.

Ratings and reviews

Does not work with my desktop server, especially when the prompt requires a web search. It ends up in an infinite loop, searching for variations of what it needs, forever. Please fix! Also, it does not work with certain models like Qwen 3.5 9B. It complains about 1024 minimum image tokens, and quits. Increasing the context window in settings does not help.

I really want this app to work, but as others have said it still has a lot of problems that need to be addressed. There seems to be no way to link it to the desktop AnythingLLM even though its gui claims this is supported. Connecting to my Ollama server worked but it enters infinite loops on searches and times out so much (especially on the initial query) that the app is basically unusable for anything I installed it for. I'd love to try this again later after it has had more work.

The workspace sync is great if you already have a local anythingllm setup. It's giving me access to all my tools I have already setup in one go. Not sure why so many have issues with the setup. For me connecting to my ollama was just entering the address and selecting model. For workspace sync with local anythingllm it was only scanning a QR code.

What’s new

What's new in 1.2.0:
• Memories: save facts globally or per workspace
• Scheduled jobs: run prompts on a schedule
• "Ask with AnythingLLM" in the text-selection menu of any app
• Share images, docs & links into AnythingLLM
• Read Word, Excel, PowerPoint & more; create docs, PDFs & slides
• Model fit badges + Hugging Face "Use this model" support
• Get notified when a reply finishes while your phone is locked
• Better voice input, faster replies, redesigned new workspace flow & bug fixes