The Android Harness

@AndroidHarness

A Native Android AI Coding Agent harness app.

16 Following5 Followers46 Posts
  1. The Android Harness@AndroidHarness
    This is seriously impressive. Just ran @liquidai's LFM 3 1.7B (F16) locally on my Android phone, right inside AndroidHarness's Chat section, and it works like a charm. #AndroidHarness #LiquidAI #LocalAI #OnDeviceAI
    Download
    Liquid AI @liquidai
    Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵
    00294
  2. Liquid AI✓@liquidai
    Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵
    Download
    5031.2K8.9K2.2M
  3. The Android Harness@AndroidHarness
    Shipped AndroidHarness 2.9. Your phone can now sync projects via FTP/SFTP, share workspaces over FTP, rotate API keys across AI providers, and use custom proxies. Plus 5 new languages. Real development, directly on Android. Open source. #AndroidDev #OpenSource #AIAgents
    Download
    10256
  4. The Android Harness@AndroidHarness
    I just built an Android app using an Android app. Sounds ridiculous, but watch this. AndroidHarness wrote the code, compiled the APK, and installed the finished app directly on my phone. No PC. No Android Studio. No cloud build service. Actual Android development, entirely on Android. #AndroidDev #AICoding
    Download
    00356
  5. The Android Harness@AndroidHarness
    Just shipped Android Harness 2.7! I've been working on something I've wanted in the app for a while: a proper Chat section alongside the coding agent. You can now just chat, research, brainstorm, or ask questions without opening a coding workspace. It has automatic web + GitHub tools, deep research with up to 100 sources, rich replies, attachments, and shared memory. And my favourite part? If a conversation turns into something you want to build, you can take that context straight into Code and continue from there. All on Android. No PC needed. #AndroidHarness #OpenSource #AndroidDev #AI #AIAgents #Coding #BuildInPublic
    Download
    7010295
  6. The Android Harness@AndroidHarness
    Asked @deepseek_ai v4.1 flash 1017 from @runinfrai to make a website about our AndroidHarness using AndroidHarness on phone. The results are really good and clearly an improvement from DS4.1 Here is the results that we got. We should test the new @ornith_ model on RunInfra Next and compare. #OpenSource
    Download
    RunInfra (YC F26) @runinfrai
    DeepSeek V4.1 Flash 1017 is live on RunInfra hour 0 support. day 0 is for slow teams 327 tok/s output. 30ms to first reasoning token $0.15 per 1M input, $0.60 per 1M output, $0.01 per 1M cached. 1M token context cache hit rate stays above 98% on agentic workloads. every hit and miss shows up in your dashboard, so you can check it yourself OpenAI-compatible chat completions and Anthropic-compatible /v1/messages. text and image input zero data retention. never used for training
    10263
  7. RunInfra (YC F26)✓@runinfrai
    DeepSeek V4.1 Flash 1017 is live on RunInfra hour 0 support. day 0 is for slow teams 327 tok/s output. 30ms to first reasoning token $0.15 per 1M input, $0.60 per 1M output, $0.01 per 1M cached. 1M token context cache hit rate stays above 98% on agentic workloads. every hit and miss shows up in your dashboard, so you can check it yourself OpenAI-compatible chat completions and Anthropic-compatible /v1/messages. text and image input zero data retention. never used for training
    Download
    14919218.8K
  8. The Android Harness@AndroidHarness
    After I use Abliterated Model GLM 5.3 by @abliteration_ai to get an uncensored answer, and then if I continue the same chat with deepseek or mimo models or any opensource models, It actually continues doing the same task which Abliterated-model-large-v2 did without refusal. Interesting 🤔. #AI #GLM53 #AbliteratedAI #DeepSeek #MiMo #OpenSourceAI #UncensoredAI #LLM #AIResearch #ModelAlignment #AIExperiment #PromptEngineering
    Abliteration @abliteration_ai
    We're excited to announce abliterated-model-large-v2 (GLM 5.3 abliterated) is live on @AskVenice!
    003165
  9. Abliteration@abliteration_ai
    We're excited to announce abliterated-model-large-v2 (GLM 5.3 abliterated) is live on @AskVenice!
    192443132.3K
  10. The Android Harness@AndroidHarness
    Been testing DeepSeek V4.1 Flash from @runinfrai inside AndroidHarness on my phone. Pretty impressive so far, especially the 97.4% overall cache hit rate! #AndroidHarness #DeepSeek #AIAgents #OpenSource
    Download
    RunInfra (YC F26) @runinfrai
    deepseek V4.1 flash is live on runinfra 378 tok/s output. 30ms to first token $0.14 / 1M input. $0.03 cached. $0.58 output 1M context. 99.3% cache hit rate over the last 24h OpenAI-compatible chat completions and Anthropic-compatible /v1/messages -- drop it into claude code as is zero data retention by default
    003188
  11. RunInfra (YC F26)✓@runinfrai
    deepseek V4.1 flash is live on runinfra 378 tok/s output. 30ms to first token $0.14 / 1M input. $0.03 cached. $0.58 output 1M context. 99.3% cache hit rate over the last 24h OpenAI-compatible chat completions and Anthropic-compatible /v1/messages -- drop it into claude code as is zero data retention by default
    341635739.7K
  12. The Android Harness@AndroidHarness
    Why remotely control a coding agent on your computer when you can run one directly on your phone? No virtual environments, no remote machines, no laptop required. While @cursor_ai brings desktop agents to mobile, AndroidHarness brings the entire AI coding experience natively to Android. Completely free and open source. GitHub: Play Store: #AndroidHarness #OpenSource #AIAgents #AndroidDev #VibeCoding
    Cursor @cursor_ai
    You can now control agents on your computer from your phone. Check in, reply, or start new tasks from the Cursor iOS app.
    10191
  13. Cursor✓@cursor_ai
    You can now control agents on your computer from your phone. Check in, reply, or start new tasks from the Cursor iOS app.
    Download
    3001552.3K179K
  14. The Android Harness@AndroidHarness
    You can now natively use ChatGPT subscription for coding on your phone through Android Harness without any PC or Root. New Android Harness 1.4 update brings you the option to use your ChatGPT subscription. Along with local models update, install and run GGUF and Safetensors AI models right from your phone for coding with vision support and tool calling. This update also ships the feature to use a different AI model for SubAgents. You can choose a different provider and a different AI model for SubAgents. Install Android Harness on GitHub or from Play Store #AndroidDev #AndroidDevelopment #AICoding #AlAgent #CodingAgent #OpenSource #Kotlin #JetpackCompose #DeveloperTools #MobileDevelopment #GitHub #Programming #SoftwareDevelopment #Al
    Download
    Tibo @thsottiaux
    You can now use your ChatGPT subscription directly in over 16 partners products. No little rules, you can just use all your included usage right there. This includes things such as Devin, OpenCode, Notion, and many more. Sign in with ChatGPT and go.
    003199
  15. Tibo✓@thsottiaux
    You can now use your ChatGPT subscription directly in over 16 partners products. No little rules, you can just use all your included usage right there. This includes things such as Devin, OpenCode, Notion, and many more. Sign in with ChatGPT and go.
    5242759.3K874K
  16. The Android Harness@AndroidHarness
    Finally, Android Harness has been approved by the Google Play Store. An Ai Coding Agent On Mobile No Root No PC. You can do pretty much anything that an AI agent harness can do on desktop, directly from your Android phone. AndroidHarness is a real AI coding agent that can work directly on your projects without needing a PC. It can read and edit files, run terminal commands, use Git and GitHub, build and test projects, search and browse the web, preview websites, connect to remote machines over SSH, and work through a full AI chat interface. Some of the main features include: Full AI coding agent with tool calling, built-in code editor, file manager, visual diffs, multi-file editing, patch support, Git and GitHub integration, terminal access, background processes, Build & Test dashboard, web search, browser automation, localhost and website preview. Remote development over SSH and SFTP, MCP server support, CodeGraph integration, repository maps, parallel subagents, custom skills, memory, todo tools, scheduled AI automations, voice input, Android notifications for permissions and agent questions. Runs can continue while your screen is off, interrupted tasks can be resumed, file changes can be reverted selectively, permissions can be controlled per tool, token, cost and time limits can be configured, context can be compacted without deleting your chat history. It supports Anthropic, Google Gemini, OpenAI-compatible APIs, custom API endpoints and model IDs, built-in keyless free models, multiple thinking levels, model latency and cost tracking, and dual-model workflows where one model can plan while another executes. The goal is simple, bring the Claude Code, OpenCode and desktop AI coding-agent experience directly to Android. No laptop required, just your phone and your code. Open source, MIT licensed. Google Play: GitHub: AndroidHarness also wouldn't exist in its current form without the incredible open-source projects and developers that inspired parts of it. Huge thanks to @NousResearch for Hermes Agent, @paulgauthier for Aider, @pidotdev for Pi, @opencode for OpenCode, @claudeai for Claude Code, @roocode for Roo Code, @cline for Cline, @browser_use for Browser Use, @termuxdevs for Termux, the developers behind Shizuku, @getcodegraph for CodeGraph, @JuliusBrussee for Caveman, @ggml_org for llama.cpp, @Rosemoe188 for sora-editor, and the developers behind Eruda. AndroidHarness borrows ideas and design inspiration from across this open-source ecosystem. Massive respect to everyone building these tools in the open. #AndroidHarness #AndroidDev #AndroidDevelopment #AICoding #AIAgent #CodingAgent #OpenSource #Kotlin #JetpackCompose #DeveloperTools #MobileDevelopment #GitHub #Programming #SoftwareDevelopment #AI
    Download
    301552
  17. The Android Harness@AndroidHarness
    Native AI Coding Agent in your pocket. No root no pc Here are some tests that I ran on it. Video 1: Dual planning mode one for planning and one for executing Video 2: Browser Control. screenshots. Debug it. Video 3: install Linux packages by agent / manual Video 4: build Android app and debug it using vision. Has been an interesting journey and very intense to make it available on mobile. #buildinpublic #opensource #androiddev #Alcoding All possible thanks to some of the greatest projects that Android Harness grew along, from @NousResearch @opencode @cline @pidotdev
    Download
    00385
  18. The Android Harness@AndroidHarness
    Vibecode on your phone without pc. I made a native android app for vibecoding and it's fully opensource and under MIT. Can do almost everything Has linux environment inside (need to install inside app) Elevated with shizuku (optional) No Root Needed Has a live catalog to pull ai providers. GitHub signing using access tokens Web search with BRAVE search and TAVILY api keys MCP servers (local stdio/HTTP/SSE) support Memory support Skills support Has stats Pretty much everything works as of now. Check the GitHub to get the latest apk Or go to actions workflows section to get a nightly build
    001116
  19. The Android Harness@AndroidHarness
    Android Harness now supports full GitHub Actions and workflow integration. Connect your account using a GitHub personal access token to access and edit all of your projects directly from the app. This release also includes several critical bug fixes and security improvements. Get the Android Harness app on GitHub: #ai #harness #aiagent #coding #android #mobilecoding #codingagent
    Download
    00147