SOVEREIGN ENGINEERING RUNTIME / v0.1 POC

Autonomous software
engineering, under your control.

AutoSE turns a plain-English requirement into reviewed, governed code. A complexity classifier routes each request to the right-sized pipeline — from a quick single-agent fix to a full design → plan → execute → validate loop with ADRs — all on your machine. No cloud. Nothing leaves your infrastructure.

Download for Windows Desktop app · .exe installer Get the CLI macOS · Linux · Windows

The CLI: one line, any OS

$ curl -fsSL https://autose.dev/install.sh | sh

Needs git; the installer sets up uv and Python for you. Bring any OpenAI-compatible endpoint like Ollama.

Runs offline Org-grounded Human-in-the-loop Small local models Made in India

TRY IT

Run a session, right here.

Pick a requirement and watch AutoSE read the project, sketch a plan, do the work, and file its report, the same worklog you'd see in the app. Stop it mid-run if you like. The CLI gives you this exact runtime in any terminal, on any OS.

AutoSE
AutoSE
New requirement
Describe a feature, a bug, or a question about your project. Plain English is perfect.
ENTER ↵ STARTS · SHIFT+ENTER FOR A NEW LINE
Ready MODEL qwen3-coder:30b LOCAL / NO CLOUD

THE WORKFLOW

One request flows through the right-sized pipeline.

A classifier reads your prompt and automatically scales the pipeline to match — simple asks stay fast and light, complex ones get deeper design and validation, including a loop that sends work back for another pass before it ever reaches you.

  1. PH-00

    Classify

    A single call sizes up the request and allocates just enough pipeline for it.

  2. PH-01

    Design

    Architecture is analysed against your ADRs and governance rules.

  3. PH-02

    Plan

    A detailed, ordered implementation plan is produced from the design.

  4. PH-03

    Execute

    Every step of the plan is implemented against your codebase.

  5. PH-04

    Validate

    Tests and lint run automatically; failures loop back for another pass.

  6. GATE

    Your review

    Inspect, diff, approve, or reject. Nothing writes without sign-off.

SMART ROUTING

Right-sized effort for every request.

Not every change needs an architecture review. AutoSE classifies each prompt behind the scenes and only spends as much reasoning — and as much compute — as the task actually needs. You never pick a mode; it just fits the work.

SMALL ASKS

Quick, focused fixes

Small edits, typos, and one-off tweaks are handled directly — no plan, no ceremony, just the change.

EVERYDAY WORK

Planned, tested changes

Typical features get a drafted plan, the code written against it, and tests added and run before you see the result.

BIGGER CHANGES

Designed, validated, looped*

Larger or riskier work gets architectural analysis, ADR generation, and a time-boxed validation loop that retries failed steps before handing off to you.

* In active development and rolling out as a premium tier.

CAPABILITIES

Engineering you can account for.

SOVEREIGNTY

Runs on what you own

All inference goes through Ollama on your machine or an org-hosted endpoint. It calls no third-party APIs and sends no telemetry.

CONTEXT

Grounded in your org

A profile encodes your coding standards, approved libraries, and architectural rules. Every output is shaped by them.

GOVERNANCE

Traceable by default

Requirements link to decisions, decisions to code, and code to review. The audit trail is part of how the runtime works.

CONTROL

You hold the pen

Human review is a built-in step. Agents prepare and propose. You approve, refine, or reject.

ROUTING

Small local models, right-sized

Work is routed between smaller and larger local models by complexity, so most tasks run on small, efficient models instead of an oversized general-purpose one.

SURFACE

Interactive TUI or desktop app

Drive workflows from an interactive terminal session, or from the AutoSE desktop app — same local-first core underneath either surface.

SUSTAINABILITY

Lighter on energy and carbon

Small local models sized to the task draw far less power than routing every request to a giant cloud model, cutting the energy and carbon cost of everyday engineering work.

CODE HEALTH

Debt-aware, not debt-adding

Every generation is checked against your organization's coding standards and architectural constraints, so output reduces technical debt instead of quietly adding to it.

GET STARTED

Three commands to your first workflow.

On Windows, the desktop app is one download away. Everywhere (macOS, Linux, Windows) the CLI below gets you the same runtime.

Download the Windows app (.exe) Prefer the terminal? Follow the steps below.

  1. 1
    Install the CLI
    Run the one-line command for your OS (in the panel). It clones AutoSE, sets up its environment with uv, and adds the autose launcher to your PATH.
  2. 2
    Set up a local model
    AutoSE runs on inference you own. Install Ollama and pull a coder model: ollama pull qwen2.5-coder:14b
  3. 3
    Run it in your project
    From any project folder, run autose for an interactive session, or autose "add tests for the parser" to start from a prompt.
macOS / Linux 
$ curl -fsSL https://autose.dev/install.sh | sh
Windows (PowerShell) 
> irm https://autose.dev/install.ps1 | iex

Want to read it first? Both scripts are plain text. See install.sh and install.ps1. You can also build from source by following the README.