Oystro keeps your AI coding on track.
Oystro is an Agentic Development Lifecycle (ADLC) protocol for your AI coding agent. It provides the guardrails for one simple process: agree what to build, plan it, build it, and check the proof before it ships.
The agent does the work and stops twice for you — approve the plan before it starts, approve the proof before it goes live. Everything in between, it handles alone.
Oystro keeps the story of the project in one place: what was built, why, and what proved it. Any agent can pick it up where the last one stopped, and you always stay in control.
Oystro OSS is free and open source today. Oystro Standard, the enhanced version, is on its way — join the mailing list to be notified.
The change Oystro is bringing.
A typical coding session produces AI slop: you prompt the agent to build an app, or to analyze existing code and fix an issue or change a feature. The agent has no consistent way of doing it, and no history is kept.
- ✕The plan lives in chat and scrolls away.
- ✕You only see the decision after the code exists.
- ✕Every session starts from zero.
- ✕The only proof is the generated code.
Oystro provides the guardrails to follow an Agentic Development Lifecycle (ADLC) and generate the project artifacts — from intent to a released, tested change. No more AI slop.
- ✓initsets up the project: goals, rules, architecture, and requirements.
- ✓defineturns the idea into one plan: what to build, the tasks, and the checks.
- ✓Human gates1. approve the plan before the build, 2. approve before the release.
- ✓Evidenceevery change is tested, and the result is tied to the commit.
- ✓Historyany agent can pick up where the last one stopped.
The protocol is live. Start today.
No need to wait for the paid launch to get value. Oystro OSS is the protocol itself, open source and running now: the workflow, templates, and rules. File-based and agent-agnostic, with no runtime, no account, and no API keys. The paid engine, Oystro Standard, adds the memory and the enforced gates that make it dependable.
During init and define, Oystro adopts a workflow size — Small (S), Medium (M), or Large (L). An S change is a quick tweak: define, build, verify, then you approve the release. An M change gets the full flow with both human gates: approve the spec before the build, approve the evidence before the release. An L change adds a fresh-eyes agent review before verification. The level sets how much ceremony a change gets, so a config tweak never waits on a schema change's process — you opt into speed; the process doesn't force it.
Oystro has predefined agent profiles, each with a bounded role: a Collaborator helps you define what to build, a Developer builds one task at a time, a Sr Tech Lead reviews the build against the plan, and a Gatekeeper assembles the evidence. None of them can approve the work. That's yours — at Gate 1 and Gate 2. You bring the intent and the approvals; the agents do the rest, and the release ships.
The short answers.
What is Oystro?
Oystro is an Agentic Development Lifecycle (ADLC) protocol for AI coding agents. It takes a change from intent to release along a set path (plan, approve, build, verify, release) and keeps the decisions, evidence, and history in one place.
How is it different from just prompting my agent?
The work is planned before it is built, and proven before it ships. The spec is a file you approve, “done” needs evidence tied to the commit, and the next session resumes from the repo instead of starting from zero.
What are the two gates?
Gate 1: you approve the plan, spec, and tasks before any code. Gate 2: you run acceptance tests and review the evidence before the release. Between them, the agent works on its own.
Which agents does it work with?
Any agent that can read and write files: Claude, Codex, Cursor, Copilot, OpenCode, Gemini, or whatever comes next. The workflow stays; the agent is interchangeable.
Does the result depend on the model?
Yes. Oystro provides the protocol; the outcome depends on your input at init and define and on the model's capability. Some models follow the protocol more strictly than others. Oystro is agent- and model-agnostic, so you can switch agents or models without changing the protocol.
Does it work on new and existing projects?
Yes. On an empty folder the agent builds the foundation from your answers; on an existing repo it reverse-engineers the architecture first. Every change then runs through the same gates.
What is the difference between Oystro OSS and Oystro Standard?
Oystro OSS is the protocol: the workflow, templates, and rules, as files in your repo. Oystro Standard is the paid engine: enforced gates, evidence capture, project memory, and a dashboard. Start free; add the engine when you want the gates enforced.
How do I install it?
Nothing to install. Open your agent in your project and paste one command. Step by step: Install Oystro in the User Guide →
Is it free?
Oystro OSS — the workflow, templates, and rules — is free and open source, licensed AGPL-3.0: file-based, with no account and no API keys. Oystro Standard, the paid engine, launches in December 2026; waitlist members get first access.
Is Oystro production-ready?
Oystro OSS is live today and ready to run on your projects: the workflow, templates, and gates are free and usable now. Oystro Standard ships in December 2026. Try the OSS on your own code and judge it there.
Will it make my project more maintainable?
Yes. Every change is one commit tied to an approved spec, and the build narrative records what changed, why, and what proved it. Months later, you can still explain why the code works the way it does.
Notes on harness engineering.
The ideas behind Oystro, written up in full.
Every agent session starts from zero
Why long projects lose the thread, and why the fix is a process, not a better model.
Harness engineering vs vibe coding
The difference isn't speed. It's who's in the driving seat.
Why the harness matters more than the model
The model gets you speed. The harness gets you trust.
The five levels of AI-assisted development
Most teams think they're further along than they are.
What a human gate looks like
A concrete walkthrough: approve the plan, build, review the evidence, release.
Why vibe coding breaks at scale
When AI writes code faster than teams can verify it, software entropy accelerates.
Hey - I am George.
We are building Oystro because AI coding agents are fast, but speed without oversight is how projects drift from what was intended. We want work you can review, audit, and stand behind: human gates, recorded decisions, and evidence with every change. We are Team Oystro — a small group that shares the same goal. Watch the demo and leave a comment (we reply there), or DM us on X.
Try Oystro OSS today, or join the waitlist.
Oystro OSS is free and live. The paid engine — enforced gates, evidence capture, and project memory — ships in December 2026; waitlist members get first access.