Agentic SDLC. Your AI ships code. You still decide.
Turn plain-language feature requests into reviewed, human-approved changes. The agent proposes. You, or your automated checks, approve. The server deploys.
Built into OSC My Apps by Jonas Birme. The agent has zero deploy authority -- every change waits behind a review gate before touching your running app.
The agent has no authority to apply its own changes.
Most AI coding tools treat deploy authority as a default. Agentic SDLC on OSC does not. Authorization is held server-side. The agent can propose, but it cannot act. That boundary is architectural, not a policy setting you can accidentally disable.
The agent works in an isolated Gitea copy of your codebase. Your live application is never touched until you choose to merge and deploy. If the agent produces something broken or off-target, you reject it -- nothing has changed.
The review gate is mandatory for every change. You can configure it as a human step or hook your CI pipeline to approve automatically when tests pass. Either way, the gate runs. The agent cannot route around it.
Every arrow is a deliberate step
The agent cannot skip any stage in the pipeline. From the moment you file a request to the moment a change reaches your running app, every transition requires an explicit handoff.
Feature request
File a plain-language feature request or bug description for any My App on OSC.
Isolated copy
The agent works inside an isolated Gitea copy of your codebase. Your live app is untouched.
Agent proposes change
The agent writes the change and opens it for review. No authority to merge. Ever.
Review gate
You approve manually, or your automated checks approve. The gate is mandatory and cannot be skipped.
Server-held auth deploys
On approval, the server-held authorization deploys the change. The agent key never touches production directly.
Feature request → Gitea isolated copy → Agent proposes → Review gate → Server-held auth → Deploy
Your Claude key. Your OpenAI key. Your bill.
OSC adds zero markup on inference. You connect your own Anthropic or OpenAI API key in the My Apps settings. The inference cost appears on your AI provider bill exactly as it would if you called the API directly -- because you are.
If a better model ships tomorrow, swap the key. You are not locked to a specific model version or a specific AI vendor. That is the same zero-lock-in principle OSC applies at the infrastructure layer, extended to the AI layer.
Turn it on in any My App.
Agentic SDLC is available on every paid OSC plan. There is no separate product, no migration, no new infrastructure to provision. Open the settings for any My App, enable Agentic SDLC, and paste your Claude or OpenAI API key. The feature activates immediately.
Open My App settings
Any My App on your OSC account
Enable Agentic SDLC
One toggle in the settings panel
Paste your API key
Claude or OpenAI -- your choice
Frequently Asked Questions
What happens if the agent proposes a broken change?
The review gate rejects it. Nothing deploys. The agent has no authority to apply its own changes. Every proposed change waits for explicit approval from you or your automated checks before it touches your running app.
Can I use my own automated checks as the approver?
Yes. Hook your CI pipeline to the review gate. If your tests pass, the change is approved automatically. If they fail, the change is held for human review. You define the approval criteria.
Do I need a GitHub account?
No. OSC includes built-in Gitea hosting. Create repositories directly on OSC -- no external git account required. You can also connect GitHub or any HTTPS git host if you prefer.
What models does the agent support?
Claude (Anthropic) and OpenAI models. You bring the API key. You control which model runs, and you can swap to a better model any time one becomes available. No vendor dependency at the AI layer.
Does Eyevinn see my code?
Your code stays in your workspace. The agent operates inside an isolated codebase copy (a separate Gitea environment) that does not affect your live app until you approve and deploy.
How much does Agentic SDLC cost on top of my plan?
Your OSC plan covers the platform. Inference cost is whatever you pay Anthropic or OpenAI directly. OSC adds zero markup on AI inference. The BYO-key model means you see the actual API cost, nothing more.
Ship code with confidence
Start with a free OSC account. Enable Agentic SDLC in any My App settings and bring your own Claude or OpenAI API key. The review gate is always on.