Most AI workflows vanish with the tab that launched them. Zorai gives you a shared, watchful runtime where agents keep working, memory persists, approvals stay visible, and every surface reconnects to the same live state.
Zorai turns scattered prompts, scripts, tabs, and terminal state into one operating system for agent work. The runtime stays awake, so the work does too.
Give an agent a real objective and let the daemon hold the state while you move between surfaces.
Approvals, sandboxing, and policy checks make risky actions legible instead of magical.
Shells, panes, and sessions remain native tools — but they now live inside a broader agent runtime.
Inspect what the agent did, which tools it used, and how the run actually unfolded.
Capture checkpoints before risky changes and roll back with evidence when needed.
Reopen prior runs, replay transcripts, and recover the operational context that usually gets lost.
Arrange terminals, agents, and panels around the workflow instead of bending the workflow to the UI.
Zorai starts with three named agents: one to do the work, one to coordinate the flow, and one to watch the edges. Together they make agent execution operable, not theatrical.
The main working fire. Svarog plans, executes, delegates, remembers, and keeps durable goals moving.
The concierge fire. Rarog carries context, smooths handoffs, and keeps coordination lightweight.
The governance fire. WELES keeps autonomy bounded, reviewable, and safe when work reaches a dangerous edge.
Operators work through Svarog. Svarog hands off coordination to Rarog and routes safety-critical work through WELES before execution.
Electron, TUI, CLI, MCP clients, and chat gateways are all just ways of touching the same daemon. Close one surface, open another, and the work is still there.
Execution, delegation, collaboration, and durable goal runs all happen inside the same live runtime.
Identity, reusable facts, and operator context accumulate instead of evaporating between sessions.
Governance is visible: approvals, isolation, validation, and auditability are part of the workflow.
Email, chat, calendars, and search enter the runtime as tools your agents can actually use.
Useful agentic work needs continuity. Zorai keeps identity, reusable knowledge, and operator preferences in curated layers instead of letting context reset every session.
Agent identity file. Defines core persona, operational constraints, and communication style. Loaded at startup — shapes every agent response.
role: artisan-craftsman
cadence: solar-cycle
personality: strict but fair
Curated episodic and semantic memory. Stores workspace facts, project conventions, recurring corrections, and durable preferences.
workspace: ~/gitlab/it/zorai
lsp: rust-analyzer
operator: conservative risk
Operator profile. Backed by SQLite — language preference, CV path, interests, session rhythm, and known feedback patterns.
language: English
session: ~11:00 UTC
signal: intellectually rigorous
Agent autonomy without visible control is just a demo. WELES makes risky execution reviewable, bounded, and auditable before work crosses a dangerous line.
WELES intercepts shell commands and risky tool calls before execution. Destructive patterns, credential exfiltration, and unsafe escalation paths are blocked or reviewed.
High-risk execution can run inside sandboxed boundaries with visible scope. The runtime makes blast radius legible before work crosses sensitive edges.
High-risk operations trigger human-in-the-loop approval. WELES presents blast-radius warnings and waits for operator confirmation.
All governance decisions are written to a WORM (Write Once Read Many) ledger. Immutable audit trail for compliance and forensics.
Email, calendars, chat, and web search become governed tools your agents can use without leaving the same execution environment.
Read inbox, search, send, trash, archive, star, mark read/unread, and view full threads.
List events, view details with attendees, create, update, delete, and manage multiple calendars.
Send messages to channels. Operators get notified when agents complete long-running tasks or need input.
Send to channels or direct messages. Human-in-the-loop approvals routed through Discord when configured.
Send to configured chat. Real-time agent notifications and approval requests via Telegram bot.
Agents search the web for real-time information. Results fed directly into agent reasoning context.
Use frontier models for strategy, fast models for implementation, and local models when privacy matters. Switch providers without rebuilding the workflow around a single vendor.
Install the runtime in seconds, or run the daemon and UI surfaces from source when you want the full development loop.
zorai --help right after install to verify it worked# Install the latest Zorai release
curl -fsSL https://raw.githubusercontent.com/mkurman/zorai/main/scripts/install.sh | sh
# Verify the install / show available commands
zorai --help# Install via npm (all platforms)
npm install -g zo-rai# Linux
curl -fsSLO https://github.com/mkurman/zorai/releases/latest/download/zorai-linux-x86_64.zip
# macOS Intel
curl -fsSLO https://github.com/mkurman/zorai/releases/latest/download/zorai-darwin-x86_64.zip
# macOS Silicon
curl -fsSLO https://github.com/mkurman/zorai/releases/latest/download/zorai-darwin-arm64.zip
# Windows
curl -fsSLO https://github.com/mkurman/zorai/releases/latest/download/zorai-windows-x64.zip# Clone the repository
git clone https://github.com/mkurman/zorai.git
cd zorai
# Start the daemon
cargo run --release --bin zorai-daemon
# In another terminal: launch the TUI
cargo run --release --bin zorai-tuiA single runtime rendered as terminals, agent panels, execution graphs, and searchable history.
Install Zorai, point it at real work, and keep the runtime alive even when the window closes.