Open Source CLI Agent

◉ ARIA

Autonomous Reasoning and Intelligent Agent

Works inside your project boundaries.
Reason before action.

Get Started View on GitHub →
aria — nemotron-3-super:cloud
◉ aria(my-project) › Build a REST API with FastAPI and test it
─────────────── ◉ Plan ───────────────
1. Scaffold project with venv + git
2. Write FastAPI endpoints
3. Install dependencies in .venv
4. Run server and test
Proceed? (yes / no / modify): yes
◉ Scaffolding 'fastapi-api'... step 1
1. new_project 'fastapi-api'
◉ Installing fastapi uvicorn... step 2
2. run_command '.venv/bin/pip install -r requirements.txt'
│ Successfully installed fastapi uvicorn
◉ Running tests... step 3
│ {"status":"ok"}
─────────────── Report ─────────────────
API built and tested. Run: uvicorn main:app --reload
◉ ARIA Task complete nemotron-3-super:cloud ~/.aria/memory
Capabilities

Built different.

Not a wrapper. An architecture.

⬡

AST Pre-Validation

Code is validated at the AST level before execution. Type errors, import issues, and Python version incompatibilities — caught before they run.

No hallucination at this layer
⌖

RAG + Web Search

When stuck on an error, ARIA searches real documentation — not model memory. Stack Overflow, GitHub issues, official docs.

⟳

Loop Detection & Pivot

Same error 3 times? ARIA detects the pattern, abandons the approach, and tries a fundamentally different strategy.

◈

Plan Before Action

Every task starts with a plan. You approve it. Only then does ARIA touch your files. No surprises.

◉

Live Stream Bar

Real-time status ticker during execution. Always know what ARIA is doing — writing, installing, testing.

⬙

Project Memory

Remembers stack decisions, key choices, and project context across sessions. Every project has its own memory.

Workflow

Seven steps. Every time.

Mandatory. No shortcuts.

01

Load Context

Project memory, decisions, history

→
02

Clarify

One question max if ambiguous

→
03

Plan

Detailed steps, your approval

→
04

Execute

Step by step, live status

→
05

Validate

Tests run, failures fixed

→
06

Remember

Decisions saved to memory

→
07

Report

What was built, how to run

Commands

Everything you need.

/fixFix bugs in current project
/testRun tests, fix failures
/explainExplain any file or concept
/commitSmart git commit
/reviewCode review with suggestions
/modelSwitch model mid-session
/memoryShow project memory
/statusSession info
Install

One command.

Install
pip install aria-x
Run
aria --model nemotron-3-super:cloud

Requires Python 3.10+ · Works with any Ollama model — local or cloud

nemotron-3-super:cloud devstral-2:123b llama3.3 qwen2.5-coder:32b gemma4:31b-cloud