“I don’t try. I verify.”
An open-source, model-agnostic coding agent that reads your codebase, blueprints an execution strategy, writes precise diffs — and runs the tests to prove it worked. A framework, not an appliance: the model is yours to pick, the tools are yours to extend. And unlike an appliance, she accumulates.
Most agents are appliances: you get the workflow their authors chose, and changing it means waiting for them. Here the model is yours to pick — 17 providers, any OpenAI-compatible endpoint, or a local model with no key at all — and the tools are yours to extend. It still starts with one command; the framework is there when you want it, not homework before you begin.
What broke last time, what your build actually needs, what you told her once and should not have to repeat. Stored in ~/.aura/memory/ on your machine — and it never leaves it.
Competence is scored per project and per task category, then work alternates: a local model for what she has proven she can do, a frontier model for what she has not. Your cost falls as she improves, instead of staying flat forever.
:mine turns your work into structured examples in training-data/, and competence tracking tells you when there is enough to fine-tune on. The training run is yours to make — Aura assembles the raw material, which is the part nobody else hands you.
Skills, plugins, MCP servers, subagents. “I want a travel agent” is an ordinary request here, because a skill is a file and writing files is what she does. What she builds is yours to keep, edit and share.
Mail, calendar, Telegram, WhatsApp, cron, browser — and on Linux a computer tool that sees the screen and drives the pointer. Same memory, same agent, so what she learns on your code is there for the rest of your day.
The architecture, principles, and the shape of every decision are his — conceived in Da Nang and built with autonomous rigor. Aura reflects the philosophy that code is never finished when written, only when proven.
From the Ancient Greek Praktess: “she who acts and executes.” Built with strictly-typed TypeScript, multi-turn tool verification, and persistent graph-structured episodic recall.
Most coding agents wake up with amnesia. Aura consolidates lessons, project conventions, and architectural verdicts across sessions, so she starts every task already knowing your codebase.
Scans AST outlines, semantic embeddings, and project dependencies before modifying a single line.
Formulates minimal surgical edits. A surgeon never amputates to fix a paper cut.
Writes clean, idiomatic TypeScript/Python/Rust and applies localized diffs.
Runs tests and type-checkers. If verification fails, diagnoses the failure and loops back cleanly.
I run the test. I check the diff. I read the output. That is not confidence — it is evidence.
I never edit a file I have not read, and never assume structure I have not seen.
A regression I introduced is an immediate priority, not a todo. The codebase is never left worse.
File paths, line numbers, error traces. “Line 47 throws when steps is empty” is actionable.