agent online | v0.1.0 | open source

Your agentic workstation

An AI layer on top of your OS. Sandboxed, permission-aware, and fully local.

Built in Rust — 60ms cold start, 32MB core, zero GC.

Plan a weekend trip to Barcelona, 2 people, €800 budget
Let me find the best options...
WebSearch "Barcelona flights Mar 2026"
WebSearch "top things to do Barcelona"
Here's your weekend plan:
· Flights — Vueling, €120pp roundtrip
· Hotel — Gothic Quarter, €85/night
· Activities — Sagrada Familia, La Boqueria, beach
Total: ~€690. €110 under budget.
Fix the auth middleware — tokens aren't expiring
Let me look at the middleware...
ReadFile src/middleware/auth.rs:42-68
42 fn validate_token(token: &Token) -> Result {
43   let claims = decode(token)?;
EditFile src/middleware/auth.rs
+ if token.exp < now { return Err(Expired) }
Added expiry check. Tokens where exp < now are rejected.
Check my emails and schedule a team standup for Friday
Checking your inbox...
GmailSearch is:unread → 8 results
8 unread. 2 need attention:
· CTO — DB migration approval (urgent)
· Sarah — Q1 report, deadline Friday
3 others: newsletter, promo, GitHub.
CalendarCreate Team standup · Fri 10am
Meeting scheduled for Fri 10am. Invites sent.
Clean up my downloads folder — it's a mess
Bash
allow deny
find ~/Downloads -type f | sort && mkdir -p ~/Downloads/docs,images,videos
Bash exit 0
Scanning 127 files...
Found 8 duplicates (340MB)
Organized → /docs /images /videos. Done.
127 files sorted. Removed 8 duplicates, freed 340MB.
What would you like to build?
$ curl -fsSL https://get.lukan.ai/install.sh | bash
capabilities

Built for Real Work

Everything you need to code, execute tasks, browse the web, and automate operations.

01 //

Smart Context

Understands your project, tools, and ongoing tasks. Reads code, config, and history to give grounded, context-aware actions. Tree-sitter parsing, project memory, and incremental indexing.

project_scan memory tree_sitter
02 //

Task Execution

Run commands, edit files, call APIs, and complete multi-step tasks directly.

bash file_edit git
03 //

Browser Actions

Navigate pages, fill forms, and interact with web tools via a real Chromium instance.

chromium screenshot click
04 //

Integrations

Email, calendar, docs, sheets, databases, and cloud services in one conversation.

gmail google_docs calendar
05 //

Secure by Design

Permission-aware actions, sandboxed execution, sensitive file detection, and clear confirmations. Every tool call requires explicit approval in strict mode.

sandbox permissions .env_guard
interfaces

One Agent,
Every
Interface

Terminal UI
Lukan Terminal UI
Lukan Desktop App Lukan Web UI Lukan Embedded Terminal Lukan Remote Access

Terminal UI

Full ratatui interface with markdown rendering, syntax highlighting, session tabs, and model selector.

configuration

Your
Config

~/.config/lukan/config.json | providers
lukan-cloud active soon

Managed proxy — all models, one API key, usage dashboard

anthropic

claude-opus-4.6, claude-sonnet-4.6, claude-haiku-4.5

openai

gpt-5.4, gpt-5.4-pro, gpt-5.2-codex

google

gemini-3.1-pro, gemini-3.1-flash, gemini-3-deep-think

github-copilot

OAuth Device Flow — no API key needed

fireworks

Llama, Mixtral, DeepSeek — inference API

nebius

Cloud GPU inference — open models

ollama

Local models — fully offline, no API key

zai

Zai inference platform

+ openai-compatible

Any endpoint — vLLM, LM Studio, Together, Groq, ...

$ lukan setup → configure API keys & providers
workers

Autonomous
background
workers

Cron-scheduled agents that run while you sleep.

lukan workers
Lukan background workers
schedule 0 */6 * * * 6h
last run passed 2m 14s ago
cron scheduler
run history
per-worker tools
sandboxed execution
custom prompts
compatibility

Universal
Skills

Same format, zero migration. Works with every major agent.

compatible agents
Claude Code Anthropic
OpenClaw Open Source
Codex CLI OpenAI
Aider Open Source
Cursor Agent Cursor
Windsurf Codeium
Any agent using markdown skill files with frontmatter.
.lukan/skills/
.
├── git-commit/
│ ├── SKILL.md ← instructions
│ └── hooks.sh
├── review-pr/
│ └── SKILL.md
├── deploy/
│ ├── SKILL.md
│ └── templates/
├── daily-report/
│ └── SKILL.md
└── monitor/
├── SKILL.md
└── config.yaml
$ you: "deploy to staging" → agent loads deploy skill
your ai workstation

One AI, everywhere
WhatsApp · Telegram · Slack · Discord

Code, office, personal — lukan handles it all. Connect it to WhatsApp, Telegram, Slack, or Discord and run your entire workstation from a single chat: emails, calendar, deployments, documents, and more.

Development
Push commits, review PRs, deploy & monitor servers
Office & productivity
Emails, calendar, documents, spreadsheets
Messaging & automation
Auto-reply across WhatsApp, Telegram, Slack & Discord
Personal assistant
Reminders, notes, quick lookups, daily briefings
lukan
online
What can I help you with?
Check my emails
3 unread:
· John — Project update
· Sarah — Meeting 2pm
· GitHub — PR #42 merged
Deploy latest to prod
Deploying...
✓ deployed successfully
Summarize today's agenda
Today:
· 10am — Standup
· 2pm — Meeting w/ Sarah
· 5pm — Deadline: Q1 report
install

One command.

That's it.

$ curl -fsSL https://get.lukan.ai/install.sh | bash
Linux & macOS · WSL/Windows · Single binary · No dependencies · Open source