turri-oss
turri is an open-source coding agent intended to give engineers back control over software development through regenerative software. Users define durable system intent in ***plain; turri uses that intent and conformance tests to regenerate code rather than treating hand-maintained code as the primary source of truth.
At a Glance
- Software engineers and technical teams using AI coding agents
- Developers who need stronger control, reviewability and accountability over generated code
- Open-source developers and teams building applications from specifications
- Teams needing automated conformance and regression validation
AI Tools by turri-oss
(1)turri
Spec Driven Coding Agent CLI
Discussions
No discussions yet
Be the first to start a discussion about turri-oss
Latest News
turri 0.2.3 released with /conform for writing/running conformance tests from chat and pi package-management commands.
turri-desktop 0.2.5 released for Apple-silicon macOS and Linux; it runs turri 0.2.3 and adds whole-window themes.
Dusan Omercevic published “I stopped knowing my own software,” explaining why turri was created and its regenerative-software approach.
turri 0.2.2 released with descriptive pulse titles for regeneration work.
Products & Services
MIT-licensed command-line coding agent installable through npm. It works with pi-supported model providers, writes ***plain intent, regenerates code, runs unit and conformance checks, and exposes commands such as turri drift, turri doctor, /regenerate and /conform.
Electron desktop application for macOS Apple silicon and Linux that hosts the same turri conversation, tools, slash commands, skills and work loop in native UI, with an intent editor, live regeneration view, preview browser and terminal.
A keyboard-only browser game used as an open-source example project; its repository describes it as written in ***plain and regenerated with turri.
Market Position
turri positions itself against conventional prompt-driven coding agents and agent workflows that continually patch code, leave tests and implementation decisions to the same agent, and make code the source of truth. Its differentiator is regenerative software: durable intent and tests are reviewed by the engineer, while a harness controls the sequence of work and regenerates disposable implementation. It is complementary to pi and conceptually competes with other AI coding agents such as Claude Code, Cursor and Codex, but emphasizes intent durability and conformance rather than only code generation speed.
Leadership
Founders
Dusan Omercevic
Founder and CEO of *codeplain. Previously founded Cleanshelf in 2016, served as its CEO, later became VP Product at SAP LeanIX after the acquisition, and held product/engineering leadership roles at Zemanta/Outbrain and Najdi.si.
Žan Jonke
Founding engineer at *codeplain and the engineer credited in public launch material with building turri. Previously an AI Engineer at Solvesall, Machine Learning Engineer at Soniox, software-development team lead and data scientist/ML engineer at Munich Innovation Labs, and software developer at DFVU.
Founding Story
The project was started after its creator experienced losing understanding and accountability as coding agents rapidly generated code. The initial vision was to make critical thinking and engineering discipline part of the tool: preserve readable, reviewable intent and tests, let an agent handle code generation, and regenerate implementation when intent changes.
Business Model
Revenue Model
The turri software itself is free and open source. Users bring their own model provider/API key or sign in, and pay the provider for model usage; turri displays build costs as work runs.
Pricing Tiers
Free MIT-licensed CLI and desktop software; model-provider usage is billed separately by the provider.
Target Markets
- Software engineers and technical teams using AI coding agents
- Developers who need stronger control, reviewability and accountability over generated code
- Open-source developers and teams building applications from specifications
- Teams needing automated conformance and regression validation
- Building new applications from written system specifications
- Regenerating implementations after requirements or functional specifications change
- Maintaining software where engineers need a reviewable source of truth and reproducible behavior
- Conformance testing and release validation for agent-generated code
- Running coding-agent workflows from a terminal or desktop application
- Overnight or CI regeneration through headless operation