ArmaraOS

Your own AI control center — install in one command.

Paste the installer for your platform, then run armaraos dashboard to start the daemon and open the web UI. Powered by AI Native Lang.

By installing ArmaraOS, you agree to our ArmaraOS End User License Agreement, Terms of Service, and Privacy Policy.

Try in browser first Browse agents Read the docs

Recommended · CLI install

Run in Terminal

curl -sSfL https://ainativelang.com/install.sh | sh

Use -sSfL so redirects from ainativelang.com resolve correctly. Installs the ArmaraOS CLI and AINativeLang (ainativelang[mcp]).

Quick start (optional)

armaraos init --quick

Initialize config and open the setup wizard after the CLI is on your PATH.

Control panel in your taskbarOne click to see what your agents are doing — no browser tab or bookmark required.
Sets itself up automaticallyThe AI engine installs quietly in the background. On macOS and Windows the daemon starts, self-checks, and opens the dashboard when ready.
Run agents continuouslyDeploy standing agents and scheduled workflows that keep working while you focus on something else.
Save on recurring AI costsCompiled AI Native Lang workflows can cut token spend 2–5× on repetitive jobs compared with ad-hoc prompting.
Wire into your existing stackInstall AINativeLang into OpenClaw, Claude, Codex, or other MCP-native hosts in minutes with helper commands like ainl install openclaw.
Free to installDownload and run ArmaraOS on your machine. Bring your own model keys (OpenRouter, Anthropic, OpenAI, or local).

After install (automatic)

On macOS and Windows, the installer runs first-time setup, starts the daemon in the background, verifies it is healthy, repairs common issues if needed, and opens the dashboard in your browser. You can close the terminal when it finishes.

Optional

1. Verify setup

armaraos doctor

Checks CLI, AINativeLang MCP, config, and daemon reachability.

2. Re-open the dashboard

armaraos dashboard

Opens the web UI and starts the ArmaraOS daemon at the same time.

Upgrade later

armaraos update

AINativeLang compiler (optional)

pip install 'ainativelang[mcp]'

Python 3.10+. Scaffold, validate, and run .ainl workflows with ainl init / ainl run.

Bring your own key

We recommend OpenRouter. After you add your API key in Overview → Setup Wizard or Settings → Providers, ArmaraOS is pre-configured to use a free OpenRouter model to get you started.

Missions & continuity

Multi-step objectives that stick around

Graph-native missions with progress you can share — plus lightweight session goals even before you flip on the full mission substrate.

Graph-native Missions
Feature DAGs, worker dispatch, handoffs, assertions, stall recovery, and auto-remediation — dashboard, CLI, and REST API.
Progress, stats & share recap
Outcome % and optional effort % from tool budgets; career badges and a Share recap PNG export for social proof.
Session goals (Codex-style)
Durable multi-step objectives upsert SessionGoal graph nodes and a GOAL telemetry chip — promotion to a full mission still needs your OK.
Post-tool capability cooldown
After tools run, follow-ups like “why did you stop?” keep capabilities advertised without re-triggering conversation suppression.

CLI install

curl | sh upgrades that actually stick

v0.8.4 hardens the browser/CLI installer path — AI Native Lang layout repair, correct Settings probes, and a fresh daemon after every re-run.

AI Native Lang layout auto-repair
Install and armaraos update remove legacy shims, prefer ~/.armaraos/ainl-venv, and refresh MCP registration — no manual PATH surgery.
Daemon restart on re-install
Re-running install.sh stops the old daemon and starts a new one so the upgraded CLI binary loads — Settings and doctor see the fix immediately.
Honest AINativeLang status in Settings
CLI probe uses ainl --help + venv pip version (AINativeLang 1.x has no top-level --version) so the AINativeLang panel matches your 1.8.x venv.

Providers & desktop

More models, AI Native Lang ready on first launch

New provider paths and a desktop that bootstraps embedded AINativeLang before the daemon starts.

Ollama Cloud + local/cloud scoping
Subscriber API at ollama.com with OLLAMA_API_KEY; local Ollama stays scoped separately from cloud credentials.
OpenCode Go driver
OpenAI-compatible and Anthropic Messages paths for MiniMax and related models via OpenCode Go.
Desktop AI Native Lang prelaunch bootstrap
Embedded AINativeLang venv is installed/repaired before daemon boot — scheduled cron and MCP work out of the box.
First-run onboarding glow
Fresh-install Get started session, hardened Telegram setup wizard, and Operator terminal excerpt fallback.

Look & feel

Redesigned, easier to read, easier to talk to

A fresh dashboard for everyday use, with on-device voice when you'd rather just talk.

All-new UI/UX
Polished light + dark themes across every page, with consistent design tokens.
Talk to Armara, Armara talks back
Built-in speech-to-text and text-to-speech run on your machine — no cloud mic by default.

Agents that do real work

Premium Hands, swarms, and 40+ ways to reach them

Pre-built autonomous agents you activate with one click — and they reach you on the apps you already use.

9 token-gated premium agents
Automated browsing, deep research, predictions, video clipping, lead generation, social media, trading and more.
Multi-agent + sub-agent swarms (beta)
Agents can spawn helper agents and coordinate — every step is shown in inspectable orchestration traces.
40+ channel connectors
Talk to your agents from your phone over Telegram, WhatsApp, Discord, Slack, Signal, Email, and more.
AI Native Lang tooling that tells the truth
Invalid graphs and missing MCP adapters surface as real tool errors with repair guidance — so agents fix AI Native Lang instead of pretending a bad run succeeded.

A brain that learns

It gets sharper every time you use it

ArmaraOS quietly learns what works, what fails, and what you prefer — without sending anything to the cloud.

Graph Memory + Auto Persona Evolution
Learns your preferences from real usage signals — no questionnaires, no extra LLM calls.
Pattern Memory + Promotion
Remembers successful workflows and promotes them into reusable patterns automatically.
Auto improvement proposals
When a workflow repeats and proves itself, ArmaraOS can stage validated pattern and procedure proposals for you to review — less manual housekeeping.
Trajectory Capture
Full execution traces of every run, available for replay and review in the dashboard.
Failure Learning
Remembers and prevents repeated errors so you don't hit the same wall twice.
Smart Suggestions
Context-aware recommendations grounded in the agent's own history.
Closed-Loop Validation
Improvement proposals are validated before they're adopted — nothing changes silently.
Smart Detection
Spots recurring tasks and offers to turn them into AI Native Lang workflows with MCP tools, validation, and auto-correction.

Cost & control

Cheaper to run, easier to trust

See exactly what your agents are costing you — and cut that bill without losing context.

Settings → Vault
A credentials center for API keys and secrets: catalog dependencies, test what applies, and keep an audit-friendly record of changes.
Adaptive Compression
Learns the right amount to compress per project — up to ~70% token / API cost savings on recurring workflows.
Visual Analytics, Monitoring & Audit Logs
Dashboards for token / API spend, budget limits, system health, and a tamper-evident audit trail.

Also available

AI Native Lang & compiler

Write deterministic agents in a Python-like syntax. Compile to portable IR and deploy anywhere — cloud, edge, or locally alongside ArmaraOS.

pip install ainativelang

Explore AI Native Lang →

OpenClaw / ZeroClaw hosts

Already running OpenClaw v1.3.0+ or another MCP-native host? Wire AI Native Lang in with helper commands like ainl install openclaw and ainl status.

Easy install for agent hosts →

Desktop installers

Prefer a double-click app? Signed .dmg and .exe builds are published on GitHub Releases for Apple Silicon, Intel Mac, and Windows x64.

GitHub Releases →