01 / INTENT SPACE
◎GOALAGENT
Keep the work
in one place.
Goals, agents, evidence, and resources live together in a spatial workspace—not a trail of disconnected pages.
INTRODUCING KEPLERV
A browser for
what you
want to do.
Most browsers take you from page to page. KeplerV starts with your intention—and brings goals, research, and the work of AI agents into one space you can understand and direct.
BUILT FOR macOS TAURI · SVELTE · PYTHON OPEN SOURCE
YOUR INTENT SPACE WEDNESDAY · 10:42 AM
Plan a research tripGOAL · IN PROGRESS
AResearch agent3 SOURCES FOUND
↗Field notesWEB RESOURCE
✓Scope checkedGUARD · ALLOWED
GOAL ACTIVITYCompare field sites
Research is underway. Sources and decisions stay attached to the goal.
AGENT WORKING VIEW ACTIVITY ↗
◖ PRESS TO SPEAK VOICE READY
GOALS 01AGENTS 01ACTIVITY LIVE LOCAL WORKSPACE
01 Your goal stays at the center.
INTENT, FIRSTAGENTS WITH BOUNDARIESHUMAN CONTROL, BUILT INONE CONTINUOUS SPACE
A DIFFERENT WAY TO USE THE WEB
Less tab juggling.
More forward motion.
KeplerV is being built around the reason you went online. Give a goal, explore the work as it unfolds, and step in whenever a decision needs you.
Keep the work
02 / UNIVERSAL AGENT RUNTIME
Goal created 10:39
Public sources searched 10:40
Evidence saved with citations 10:41
Approval required YOU
Agents that show
their work.
Delegate public research to scoped agents. Follow their activity and inspect findings linked to their sources.
03 / KEPLERV GUARD
✓
REQUEST CHECKEDOUTSIDE ASSIGNED SCOPEDENIED · REVIEW AVAILABLEAutonomy with
a visible boundary.
AIP checks what an agent can do. Guard turns policy events into evidence-based investigations you can review.
K
LISTENING · PROCESSING · SPEAKING
VOICE-NATIVE, NOT VOICE-ONLY
Say what you mean.
Stay in control.
Speak a goal and KeplerV can interpret it against your workspace. Real microphone capture, transcription, and spoken responses are in the app; approvals remain deliberate human decisions.
See how it’s built ↗CONNECTED UNIVERSE · PHASE 5
Your internet,
in context.
KeplerV brings personal services into the same goal-centered environment—so you can watch, listen, read, and work while agents handle only the tasks you explicitly authorize.
Media that belongs to the moment
MEDIAOpen supported video in a focused surface, keep your workspace underneath, and return to the goal when you’re done. Local video and YouTube experiences are available in the app.
Your accounts. Your permission.
CONNECTED SERVICESConnect services through account-aware flows. Read-only email connectors support Gmail and Outlook architecture; private credentials are held in native secure storage.
Separate permissions for people and agents
HUMAN + AGENTProvider scopes set the outer limit. KeplerV’s permission layer narrows what an agent can do, records activity, and puts sensitive actions behind human approval.
One continuous workspace
EXPERIENCE SURFACESMedia, email, documents, websites, and approvals appear when relevant, then recede. Your goals and agent activity remain in view underneath.
Account-aware Guard
GUARD + AIPGuard can investigate denied requests and service-capability boundaries with cited evidence, while minimizing exposure of personal content.
Built to connect responsibly
SECURITY MODELOAuth with PKCE, scoped access, separate human and agent sessions, revocation, expiry handling, and explicit provider capability checks shape the architecture.
KeplerV is actively evolving. Provider support depends on official APIs and platform compatibility; Netflix and other DRM streaming playback are not claimed as verified.
06 THE NEXT FRONTIER · INTER-AGENT INTELLIGENCE
When agents meet,
trust needs a boundary.
The next internet will be full of agents acting for different people and organizations. KeplerV is being designed to let its agents collaborate across that boundary—with the human’s intent and permissions still in charge.
THE HUMANOwns the goal
SETS INTENTKEPLERV AGENTWorks for youSCOPED BY AIP
TRUST GATEWAYEvery exchange
has a boundaryVALIDATE · AUTHORIZE · RECORD
has a boundaryVALIDATE · AUTHORIZE · RECORD
EXTERNAL AGENTOperated elsewhereUNTRUSTED INPUT
⌕
CLAIMS MEET EVIDENCEIdentity is a signal. Behavior is still evaluated.
GUARD · PROVENANCE · VERIFICATIONIdentify advertised capabilities and distinguish verified identities from self-declared or unknown agents.
Route communication through one gateway, checked by AIP and limited to the human’s authorization.
Separate what an agent says from independent evidence, signed statements, and facts that remain unverified.
Guard can examine suspicious exchanges and cite evidence—without pretending to know another agent’s hidden intent.
✳
A verified identity is not a guarantee of trustworthy behavior.
That distinction is the foundation of KeplerV’s next phase.
A NEW CATEGORY, WITH A DIFFERENT CENTER
Not another browser
with an AI button.
KeplerV explores a different starting point: persistent intent, visible agent work, and human oversight in the same environment.
| WHAT’S DIFFERENT | KEPLERV | CHROME | FIREFOX | SAFARI | PERPLEXITY COMET |
|---|---|---|---|---|---|
| The starting point | Your goal | A website or search | A website or search | A website or search | A question to an AI |
| Work that continues | Persistent goals and agents | Mostly page by page | Mostly page by page | Mostly page by page | Conversation-led |
| Where agents work | Isolated, permission-checked runtime | Extensions and browser features | Extensions and browser features | Extensions and browser features | AI browsing session |
| Human oversight | Approvals, activity, and Guard review | Browser controls | Browser controls | Browser controls | Product-specific controls |
THE INTERNET, WITH INTENTION
Make room for
what’s next.
KeplerV is an early, open-source browser project building toward a more capable and more accountable way to work with the web.