--- title: I turned my phone into an AI agent powered by my local... description: RikkaHub Agent is a fork of the RikkaHub chat client that transforms an Android phone into an AI agent harness, offloading inference to a local LLM running on... canonical: https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: I turned my phone into an AI agent powered by my local LLM, and it installed software on its own | daily.dev og:description: RikkaHub Agent is a fork of the RikkaHub chat client that transforms an Android phone into an AI agent harness, offloading inference to a local LLM running on... og:url: https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw og:image: https://api.daily.dev/og/posts/dt5C9kQPW.png og:image:alt: I turned my phone into an AI agent powered by my local LLM, and it installed software on its own og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # I turned my phone into an AI agent powered by my local LLM, and it installed software on its own **[XDA Developers](https://daily.dev/sources/xda-developers)** · 7 min read · 2 upvotes · 0 comments ## Summary RikkaHub Agent is a fork of the RikkaHub chat client that transforms an Android phone into an AI agent harness, offloading inference to a local LLM running on a separate machine while the phone handles tool execution. The setup exposes 80+ device tools including tapping, swiping, SMS, file management, an in-app browser, SSH, Termux, scheduled jobs, and a Telegram bot. The author connected it to a Qwen 3.6 35B model running on a Lenovo ThinkStation PGX and demonstrated the agent autonomously compiling whisper.cpp in Termux, pulling a model, and browsing the web. Tools are off by default for safety, but enabling all of them balloons the context from 7,000 to 32,000 tokens. On-device inference via LiteRT-LM works but struggles with the large context window, making a remote model the practical choice for reliable tool-calling. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I run an AI agent on my Android phone that uses tools like SMS, browser control, and Termux while the LLM itself runs on a separate server? RikkaHub Agent, a fork of the RikkaHub chat client, adds over 80 device tools (tapping, swiping, SMS, an in-app browser, SSH, file management, scheduled jobs, a JavaScript skills runtime) and connects to any OpenAI-compatible endpoint, so the phone runs only the agent loop while a separate machine, such as a home server, handles inference for models like Qwen 3.6 35B A3B. _Developers wiring up local-LLM agent harnesses on mobile can track tool-calling patterns like this through daily.dev._ ### Why does a smaller on-device model like Gemma fail at tool calling compared to a larger model like Qwen when using LiteRT-LM? Gemma struggles with the same tool-calling prompts that Qwen handles successfully, likely because enabling all agent tools pushes the input context from around 7,000 tokens up to 32,000 tokens, straining prefill performance and reasoning capacity on a phone's limited hardware, whereas the larger Qwen model run on separate server hardware copes fine with the same load. _Anyone comparing local model sizes for agentic tool use can weigh trade-offs like this via daily.dev._ ### Are the numerous forks of the RikkaHub Android app safe to use? The original RikkaHub maintainer has warned that problems reported by users often come from third-party forks rather than the upstream project, and advises caution with forks generally. One known fork, discussed by its creator in a GitHub issue opened in May 2026, was approved by the maintainer as acceptable so long as it stays non-commercial and credits the original project. _Developers picking between community forks of a tool can weigh maintainer trust signals discussed on daily.dev._ ## Similar posts on daily.dev - [I turned my phone into a local LLM server, and it handles vision, voice, and tool calls](https://daily.dev/posts/i-turned-my-phone-into-a-local-llm-server-and-it-handles-vision-voice-and-tool-calls-f2fcaxaha) · XDA Developers · 0 upvotes · 0 comments - [Android AI Just Changed — and Most Developers Haven’t Noticed Yet](https://daily.dev/posts/android-ai-just-changed-and-most-developers-haven-t-noticed-yet-ew4d0vzac) · ProAndroidDev · 0 upvotes · 0 comments - [This open-source Google app turned my phone into a fully local AI lab](https://daily.dev/posts/this-open-source-google-app-turned-my-phone-into-a-fully-local-ai-lab-d9txzwke5) · XDA Developers · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#android](https://daily.dev/tags/android), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"I turned my phone into an AI agent powered by my local LLM, and it installed software on its own","url":"https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw"},"datePublished":"2026-08-07T22:00:16.579Z","dateModified":"2026-09-14T06:28:50.493Z","description":"RikkaHub Agent is a fork of the RikkaHub chat client that transforms an Android phone into an AI agent harness, offloading inference to a local LLM running on...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c23da3b17f43fe7afbe97d24ea2cfc5b?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c23da3b17f43fe7afbe97d24ea2cfc5b?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"XDA Developers","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"XDA Developers","logo":"https://media.daily.dev/image/upload/s--kCg6yyAP--/f_auto,q_auto/v1774964407/logos/xda-developers?_a=BAMAMiWQ0","url":"https://daily.dev/sources/xda-developers"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,android,local-ai","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"XDA Developers","item":"https://daily.dev/sources/xda-developers"},{"@type":"ListItem","position":3,"name":"I turned my phone into an AI agent powered by my local LLM, and it installed software on its own"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/i-turned-my-phone-into-an-ai-agent-powered-by-my-local-llm-and-it-installed-software-on-its-own-dt5c9kqpw#faq","mainEntity":[{"@type":"Question","name":"How can I run an AI agent on my Android phone that uses tools like SMS, browser control, and Termux while the LLM itself runs on a separate server?","acceptedAnswer":{"@type":"Answer","text":"RikkaHub Agent, a fork of the RikkaHub chat client, adds over 80 device tools (tapping, swiping, SMS, an in-app browser, SSH, file management, scheduled jobs, a JavaScript skills runtime) and connects to any OpenAI-compatible endpoint, so the phone runs only the agent loop while a separate machine, such as a home server, handles inference for models like Qwen 3.6 35B A3B. Developers wiring up local-LLM agent harnesses on mobile can track tool-calling patterns like this through daily.dev."}},{"@type":"Question","name":"Why does a smaller on-device model like Gemma fail at tool calling compared to a larger model like Qwen when using LiteRT-LM?","acceptedAnswer":{"@type":"Answer","text":"Gemma struggles with the same tool-calling prompts that Qwen handles successfully, likely because enabling all agent tools pushes the input context from around 7,000 tokens up to 32,000 tokens, straining prefill performance and reasoning capacity on a phone's limited hardware, whereas the larger Qwen model run on separate server hardware copes fine with the same load. Anyone comparing local model sizes for agentic tool use can weigh trade-offs like this via daily.dev."}},{"@type":"Question","name":"Are the numerous forks of the RikkaHub Android app safe to use?","acceptedAnswer":{"@type":"Answer","text":"The original RikkaHub maintainer has warned that problems reported by users often come from third-party forks rather than the upstream project, and advises caution with forks generally. One known fork, discussed by its creator in a GitHub issue opened in May 2026, was approved by the maintainer as acceptable so long as it stays non-commercial and credits the original project. Developers picking between community forks of a tool can weigh maintainer trust signals discussed on daily.dev."}}]} ```