Tin picks the best reachable route.
Ready
Tin workspace
Pulse
Your conversation, with the models and context you choose.
Model
Auto Model
Messages
0
Categories
0
Tin workspace
Tin J.A.R.V.I.S.
Voice + text AI. Tap the orb to talk, say "Hey Tin" hands-free, or type below. Tap again to interrupt.
Model—
VoiceReady
Session0 turns
Ready
Hold the orb to talk and release to send — or just tap it. You can also say "Hey Tin" to start hands-free.
Model proving ground
Compare
Send one prompt to several models at once, judge the answers blind, and let the scoreboard remember which model actually earns its keep.
Models available
0
Votes cast
0
Leader
—
Scoreboard
Deep research
Research
Ask a real question and Tin runs a multi-step investigation — planning searches, reading the top sources, and writing a single cited report.
Reports
0
Sources read
0
Searches the open web. Add a TAVILY_API_KEY or BRAVE_SEARCH_API_KEY in Vercel for higher-quality results; otherwise a keyless provider is used.
Past reports
Study desk
Homework
A cleaner workspace for assignments, class context, image help, and structured study output.
Classes
0
Model
Auto
Mode
Guide
Class-aware study buddy
Add your classes once, then ask questions or drop photos and files. Tin will figure out which class each question belongs to, calibrate explanations to your grade, and remember the patterns from each class over time.
Honest accuracy: even strong models miss multi-step problems sometimes. Tin will show its work and flag uncertain steps so you can spot errors instead of trusting blindly.
Model control
Models
Chat models, image generation, and your local library.
Installed
0
Active
Auto
Providers
0
Models / 01
Local library
Models / 02
API control center
Cloud keys, provider defaults, model routing, and external integrations now live in the dedicated API tab. This keeps the Models tab focused on local Ollama models and the app store.
Connected providers
0
Chat default
Auto Model
Models / 03
App Store · HuggingFace
Search HuggingFace's GGUF catalog. Install on PC pulls the model into your home Ollama via the bridge — works from anywhere once the tunnel is up.
FAMILY
SIZE
TASK
SORT
Loading trending GGUF models…
Tin control center
API
Manage the model providers, keys, defaults, and outside services Tin can use. Keys are stored in your encrypted Tin vault and travel through encrypted sync when sync is on.
Providers0
Ready0
CurrentAuto
Best for photos, screenshots, and step-by-step school work.
Learns in the background, then falls back when local is offline.
API / 01
Secure key vault
Current safe path
Personal keys are encrypted with your Tin passphrase on this device, then synced inside Tin's encrypted account snapshot. They are never hardcoded into frontend source files.
Future server vault
OAuth services and provider keys that should never touch the browser are shaped for a backend vault later. Until that exists, Tin labels those integrations as planned instead of pretending they are connected.
API / 02
AI providers
New here? Get Tin talking in a minute
Tin runs on local Ollama, any provider you connect below, or a built-in free fallback. The fastest start is a free OpenRouter key — paste it into the OpenRouter card below, flip it on, and Tin (voice + chat) will use it automatically.
API / 03
Model routing
Fallback mode is live: Tin tries your primary provider first, then fallback, then the rest of your enabled providers.
API / 04
Where Tin uses models
Chat default
Auto Model
Homework default
Auto vision routing
Brain local model
API / 05
External integrations
These cards reserve the real connection flow for email, calendar, repo, deploy, and community services. Anything that needs OAuth is marked planned until the backend auth flow exists.
Skills
Stacked behaviors that customize every prompt. Active skills (✓) apply to every chat. Others fire when their /trigger appears in your message — and you can stack multiple slashes in one prompt.
Total
0
Active
0
Categories
0
Private notebook
Notes
Fast capture, organized categories, autosave status, and AI transforms in one clean writing surface.
Notes
0
Category
All
Words
0
Build desk
Vibecoding
Project files, code briefs, error fixes, explanations, and build prompts in a tighter developer console.
Files
0
Mode
Prompt
Output
Ready
Vibecoding / 01
Workspace
Ready.
Vibecoding / 02
Build context
Vibecoding output will appear here once you build a prompt.
Persistent context
Tin Brain
Automation layer
Agents
Paired local workers, approved tasks, and execution outcomes for your account.
Agents
0
Backend
Local
ControlApproval required
Agents / 01
Local agent shells
OpenClaw-style local agents are prepared as a UI shell. Full control requires the local Tin backend running on your PC.