On 2026-10-13 at if(kakao)26, Kakao said it will make KakaoTalk (~50M users) an integrated agentic AI interface: Kanana Space (two-finger swipe anytime for messaging, schedules, settings; later shopping/gifting/reservations/transfers, with user approval on consequential acts); PC Kanvas for building agent teams, with a Talk entry planned H1 2027; plus Kanana-o multimodal, an AI conductor that routes by request complexity, and Project Mosaic for cross-service memory. Pre-registration opened Oct 13; Kanana Space closed beta in December.

Key Takeaways

  • ✓Official: if(kakao)26 (2026-10-13–14); tech overview Kakao Tech
  • ✓Kanana Space: summon anywhere in Talk; messaging/schedules/search/settings first, then commerce/transfers; user must approve consequential actions
  • ✓Kanvas: PC agent builder with local memory and per-agent browser sandboxes; small closed beta as early as this year; Talk hub for owned/partner/3rd-party agents targeted H1 2027
  • ✓Stack: Kanana-o omni (text/image/voice); on-device vs large via AI conductor; encoder/decoder planned open-source (not released yet)
  • ✓No public agent leaderboard yet—treat as product roadmap + demo, not GA
🧭

Turn your technical choice into a development budget

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

🔬

In-Depth Technical Analysis

Context

Messaging apps own high-frequency relationships and payments but rarely act as agent schedulers. At if(kakao)26 (2026-10-13), Kakao cast KakaoTalk (~50M users) as an integrated agentic interface—intent and execution inside existing chat/service flows, not a separate chatbot tab. Distribution stays in Talk; owned, partner, and third-party agents become a new surface, with personalization memory and payment governance as hard gates.

Architecture

Kanana Space (two-finger swipe, Ryan persona): messaging, schedules, search, settings now; later commerce/gifting/reservations/transfers, with user approval on consequential acts. Pre-registration opened Oct 13; closed beta in December. Kanvas: PC agent builder with local memory, scheduled jobs, and per-agent browser sandboxes; small closed beta as early as this year; Talk hub for owned/partner/external-framework agents targeted H1 2027. Stack: Kanana-o omni; an AI conductor routes on-device vs large models and tools by complexity; Project Mosaic shares consented memory across Talk/Space/Kanvas. Encoder/decoder are planned open-source—not yet a public repo.

Benchmarks

Public materials are keynote demos and roadmap. No reproducible agent leaderboard (SWE-bench/WebArena-class) was published with the launch. A vendor stage claim cites ~1.6× Korean tokenizer efficiency vs “competing global models” without a fixed public baseline. Treat this as product direction, not GA performance.

Developer path

Follow if.kakao.com/2026 and Kakao Tech. Watch Kanvas tool contracts and sandbox boundaries for Korea-market agents; wait for public API/SDK before assuming callable Agent APIs. Focus engineering reviews on payment confirmation UX, Mosaic consent scopes, and conductor cost routing.

Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

Up NextXpert open-sources a governed plugin platform for models, tools, skills, and agent apps
Read Now→