What's new in Snaga

The platform has grown a lot this past month — from a single AI chat into a visual command center for whole teams of agents. Here's what you can do now.

🎛️ Orchestrate — design and run AI teams on a visual canvas

Build squads visually — an n8n-style canvas: drag to connect agents, inspect any node, run/pause/activate right from the map.
Edit agents in place — rename, set schedules, switch models or knowledge bases without leaving the canvas; multi-select for bulk run/pause.
Watch work flow live — animated traces on the wires show task delegation; running and approval nodes light up in real time.
Per-node insight — runs, cost and tokens at a glance; sticky notes, auto-layout, zoom-to-icons, copy/paste, find-a-node.

🧭 Mission Control — your operations dashboard

Live fleet status — system health, spend trends, and alerts in one board.
Missions — give a goal in plain language; Snaga decomposes it into parallel objectives, runs them, and streams progress plus an after-action review. A watchdog auto-recovers stuck missions.
Approvals inline — approve or reject runs straight from the timeline.

🤖 Smarter agents

Pick a model per chat — override the LLM for one conversation without touching defaults.
Memory that builds itself — agents quietly capture lasting facts (your preferences, conventions) as you work.
Real vision — send an image, get a genuine description.
Auto-syncing model catalog — new provider models show up in the picker automatically.
Autonomy that follows you — "Remember this choice" for tool approvals now syncs across web, iOS, and your local machine.

💬 Chat & artifacts

Syntax-highlighted code, soft-wrap toggle, token estimates, Cmd+Enter to send, rename conversations.
Generative UI — interactive forms, multi-file HTML previews, file version history, and an inline Office (Word/Excel/PowerPoint) viewer.

📁 Files & knowledge

IDE-style explorer — tree plus tabbed editor, syntax highlighting, keyboard shortcuts, marquee select.
Bulk knowledge management — select-all and bulk-delete documents.

💻 Run agents on your own machine (local Snaga)

Cloud-optional — use Ollama, LM Studio, or custom endpoints with no account required.
Background workers — fire off long tasks detached, keep working, check status later; cheaper models per task.
Rock-solid bridge — reconnect circuit-breaker (no more log floods), inline streaming on Zed and Apple Terminal, styled markdown tables.

📱 iOS app — native to the core

Live Activities & Dynamic Island — run progress and approval gates on your Lock Screen.
Home-Screen widget, Control Center, quick actions, Spotlight.
Actionable approvals — Approve / Reject / Reply right from a push notification.
Offline-aware chat — clear local-agent status, reconnect toasts, per-message delivery state.
Programs — browse curricula and apply them as scheduled todos.

🎨 Design, accessibility & more

"Paper Under Glass" refresh across the app, five themes, fully responsive on mobile.
WCAG 2.2 hardening — keyboard-navigable menus, screen-reader fixes, reduced-motion and high-contrast support.
New blog with RSS and Google-News discovery (you're reading the platform now).

🎁 Try everything free

7-day Pro trial on signup — full access, no card. We'll remind you before it ends.

🎛️ Orchestrate — design and run AI teams on a visual canvas

Build squads visually — an n8n-style canvas: drag to connect agents, inspect any node, run/pause/activate right from the map.

Edit agents in place — rename, set schedules, switch models or knowledge bases without leaving the canvas; multi-select for bulk run/pause.

Watch work flow live — animated traces on the wires show task delegation; running and approval nodes light up in real time.

Per-node insight — runs, cost and tokens at a glance; sticky notes, auto-layout, zoom-to-icons, copy/paste, find-a-node.

🧭 Mission Control — your operations dashboard

Live fleet status — system health, spend trends, and alerts in one board.

Missions — give a goal in plain language; Snaga decomposes it into parallel objectives, runs them, and streams progress plus an after-action review. A watchdog auto-recovers stuck missions.

Approvals inline — approve or reject runs straight from the timeline.

🤖 Smarter agents

Pick a model per chat — override the LLM for one conversation without touching defaults.

Memory that builds itself — agents quietly capture lasting facts (your preferences, conventions) as you work.

Real vision — send an image, get a genuine description.

Auto-syncing model catalog — new provider models show up in the picker automatically.

Autonomy that follows you — "Remember this choice" for tool approvals now syncs across web, iOS, and your local machine.

💻 Run agents on your own machine (local Snaga)

Cloud-optional — use Ollama, LM Studio, or custom endpoints with no account required.

Background workers — fire off long tasks detached, keep working, check status later; cheaper models per task.

Rock-solid bridge — reconnect circuit-breaker (no more log floods), inline streaming on Zed and Apple Terminal, styled markdown tables.

📱 iOS app — native to the core

Live Activities & Dynamic Island — run progress and approval gates on your Lock Screen.

Home-Screen widget, Control Center, quick actions, Spotlight.

Actionable approvals — Approve / Reject / Reply right from a push notification.

Offline-aware chat — clear local-agent status, reconnect toasts, per-message delivery state.

Programs — browse curricula and apply them as scheduled todos.