Bedside companion
Ask about tomorrow, capture a late-night thought, and wake to the day ahead.
Open source · Native Android
Your phone,
now a little more alive.
Turn an old Android phone into an always-ready, voice-first life agent. Local-first data, BYOK intelligence, native actions.
Android 8.0+ · Bring your own AI provider · Free and open source
YouWhat's on my day?
MochiOpening your planner
Old phone. New purpose.
Not another chat app.
Not another chat app.
A companion with a place in your life.
Mochi is designed to stay nearby on a dedicated screen—ready when you speak, useful when you glance, and quiet when you do not need it.
Desk-side planner
Keep your calendar, todos, weather, and timely information one question away.
Hands-free home terminal
Wake Mochi by voice and let the conversation bring the right native screen forward.
One conversation, many capabilities
Made to understand.
Made to understand.
Built to get things done.
Mochi combines a natural voice experience with local context and typed, inspectable Tools. The result is an assistant that can move from conversation to action without becoming a black box.
Voice first
Say “Hi Mochi.”
Say “Hi Mochi.”
Then just keep talking.
Local wake-word detection and follow-up listening keep conversation flowing. Use Android speech or save multiple iFlytek and Azure Speech accounts. Choose an account, preview its voices, and opt into cloud replies separately.
Context that stays with you.
Conversation history and relevant long-term memory live on your device and return when they are useful.
Dates become a view, not just an answer.
Discuss a day or todo and Mochi can open its native calendar and planner at the right moment.
Cowork inside your private knowledge base.
Find sources, research, create documents, and edit your work through authorized Tools. Feishu includes guided self-built-app setup and account authorization, with search, wiki browsing, and document read/write access; no callback server to deploy.
Keep your accounts.
Choose your model.
Save named connections, including multiple accounts with the same provider. Official endpoint presets simplify setup; Azure and custom endpoints remain available. Supply your own API key and a tool-capable model.
OpenAIDeepSeekKimiGLMMiniMaxAgnes-AI
Keep ongoing work in view.
Keep conversation simple.
Track scheduled runs, Termux jobs, and authorized AgentLink remote work in Tasks. Open results and stop supported work without mixing ordinary chats or foreground Subagents into the list.
The optional signed Termux extension connects a separately installed Termux. Stopping an Agent does not stop work already submitted to Shell or a remote service.
Native Android Agent Browser
The web, navigated
The web, navigated
by your Android agent.
Mochi implements its own native Android Agent Browser. It can visit and read most public HTTPS websites, then navigate, click, input, and scroll to complete research directly on the page.
Explore Tools & Skills
Optional connections. Real capabilities.
Your home. Your phone. Your computer.
Your home. Your phone. Your computer.
One place to ask.
Start with Mochi for conversation, memory, and planning. Add only the connections you need: two signed extension APKs and an independent companion app.
Signed extension APK
A voice for your
Smart home · Mi Home
A voice for your
connected home.
Smart home · Mi Home
Control supported appliances and TVs, inspect device state, trigger scenes, and view the latest available camera event image in a foreground conversation.
“Turn off the living room TV.”
Connect in three steps
- Install the extension from the same release/signing channel as Mochi.
- In Tools → Extensions → Mi Home, scan the login QR with another phone signed into Mi Home; select your devices.
- Return to Mochi and choose Enable tools and Skill when ready. Installation alone does not enable access.
Unofficial connector; capabilities vary by device. Event images are not live video. Control is foreground-only; acceptance is not proof of a physical result when a device cannot report its state.
Put your phone
to work.
Run commands and scripts in Termux, process files it can access, and inspect output or stop jobs in Tasks. Once enabled, Tools can run from foreground chats, scheduled Agents, and Subagents.
ON YOUR PHONE / EXAMPLE FLOW
request → script → outputSubmit · Follow progress · Inspect resultsConnect in three steps
- Install Termux separately, plus the matching Mochi extension.
- Open Termux setup from Mochi Tools. Follow the external-command and Android permission guidance, then test the connection.
- Enable the Provider, Tools, and Skill you need. Manage submitted jobs in Tasks; disabling access blocks new calls, not existing jobs.
Full shell access under Termux permissions, not root or a sandbox. Enabled commands run without per-command confirmation. Output used by the Agent goes to your model provider.
Your coding agents.
Now within reach.
Connect Android to GitHub Copilot CLI, Claude Code, Kimi Code, Qwen Code, or DeepSeek Harness on your computer. Share a chat with your terminal; optionally grant Mochi scoped access to send work and read results.
ON YOUR COMPUTER / AGENTLINK
phone ↔ Bridge ↔ coding agentOne chat · Shared context · Your computerA separate product, not a Mochi extension APK. Requires a running computer Bridge and a signed-in coding agent. Mochi access is explicitly granted and foreground-only; remote work can continue after Mochi closes.
Mi Home and Termux extensions have no separate launcher icon: open them from Mochi Tools. AgentLink has its own app. None is required for basic Mochi use; connect and enable only what you trust.
Privacy starts locally
Your companion.
Your companion.
Your provider. Your choices.
Persona files, settings, conversations, memory, calendar, and todos stay on the device by default. Provider credentials use Android Keystore-backed storage.
- ✓ Save and explicitly switch independent AI and speech accounts
- ✓ Enable only the Tools and Skills you trust
- ✓ Location is requested only when a location Tool runs
- ✓ Review the implementation—the entire app is open source
Local-first does not mean fully offline: model requests, optional cloud speech, and enabled external Tools send necessary content to their configured services. In Settings, Share/import lets you select multiple saved connections; importing adds accounts rather than overwriting them. Anyone holding the full share link can use its included credentials and consume quota. Share it only with people you trust.
Bring Mochi home
Three steps to a more useful phone.
No Mochi account or Mochi subscription. Bring your own AI service; provider usage and optional cloud speech may incur charges.
- 1Install or update
Download v1.0.13. Choose arm64-v8a for most phones, armeabi-v7a for older ARM devices, or universal if unsure. Update through the same signing channel without uninstalling to preserve your data.
- 2Add a connection
Open Settings → AI connections → Add AI connection. Choose a preset, then enter your model and API key. Later additions do not change your current connection; select Use to switch. Switching AI cancels running local Agent tasks.
- 3Talk your way
Type or grant microphone permission to talk; enable “Hi Mochi” when ready. Android speech is built in. Optionally add a speech connection and select Use; switching speech stops the voice interaction and preview, not text or background work.
Open by nature
Built in public.
Built in public.
Better with company.
Mochi is free software under the MIT License. Explore the native Android code, shape the roadmap, report a problem, or teach Mochi a new Skill.
github.com/gongpx20069hi-mochi
A native Android, voice-first life agent—and a new purpose for an old phone.