Agentic SDLC

Agentic SDLC is a feature that attaches an AI agent to a My Apps custom application. Once enabled, the agent bootstraps a CLAUDE.md (or equivalent context file) in your repository and then runs on a regular schedule to perform autonomous software development lifecycle tasks: reviewing code, creating PRs, triaging issues, and keeping the project moving between human sessions.

What It Does

When you enable Agentic SDLC for a My App, OSC:

  1. Runs a bootstrap task that clones your repository, writes a context file (CLAUDE.md), and pushes it back to the repository via git.
  2. Creates a scheduled agent task that runs the agent on a recurring basis (e.g. daily) with access to your source code and OSC services via MCP.

You can monitor both tasks from the My Agent Tasks page.

Prerequisites

  • An OSC account on a paid plan (Personal, Professional, or Business)
  • A My Apps custom application with a Source URL pointing to a git repository
  • A git token with write access to that repository (needed for the bootstrap commit and subsequent PR creation)
  • Agent credentials configured in My Agent Tasks (an Anthropic API key or Claude OAuth token)

Enable from My Apps

  1. Go to My Apps and click on your application to open the detail panel.
  2. Locate the Agentic SDLC column (or cell) in the app row. If agent credentials are not yet configured, the toggle switch appears disabled with a tooltip: "Configure agent credentials first". Click the disabled switch to open the enable dialog, which will guide you through setting up credentials as a first step.
  3. Once credentials are configured, the toggle is enabled. Click it to open the Enable Agentic SDLC dialog.
  4. If you have configured credentials for both Claude (Anthropic) and Codex (OpenAI) in My Agent Tasks, the dialog shows a radio group to select the agent runtime: "Claude (Anthropic)" or "Codex (OpenAI)". Choose the runtime you want to use. If only one credential type is configured, the runtime is selected automatically and no radio group appears.
  5. A dialog field asks for a git token with write access to your repository. Enter the token and click Enable.
  6. The cell changes to Setting up... with a spinner while the bootstrap task runs. Once the bootstrap task completes successfully, the status changes to Active.

If the bootstrap task fails, the cell shows a Failed chip. You can re-enable after fixing the underlying issue (typically a missing or invalid git token, or a git push failure).

Reviewing and Acting on Requests

Once Agentic SDLC is active, the agent creates tickets in a Gitea repository that OSC manages on your behalf. You can review and act on those tickets directly from the OSC dashboard without needing to open Gitea.

The Requests Tab

When Agentic SDLC is active for an app, a View SDLC requests shortcut button (git pull-request icon) appears directly in the My Apps list row for that app. Clicking it navigates straight to the requests view without opening the detail panel first.

You can also open the detail panel for your app in My Apps and switch to the Requests tab. Each row represents one SDLC request submitted by you or the agent.

Click any row to open the Request Detail Drawer, which shows:

  • The request title and current status badge
  • A link to open the ticket directly in Gitea
  • Comments left by the agent (rendered as Markdown), including the agent's implementation plan
  • Sub-tasks with their individual statuses and links to the pull requests created by the agent

Approving, Revising, or Cancelling a Planned Request

After the triage agent analyses a new request it sets the status to planned and posts a detailed implementation plan as a comment. At this point the drawer shows three action buttons:

Button What it does
Approve Gives the agent the go-ahead to implement the plan. The agent starts creating sub-tasks and opening pull requests immediately.
Revise Opens a text field where you can explain what you want changed in the plan. The agent re-triages with your feedback and produces a revised plan for you to review again.
Cancel Cancels the request. The agent closes the ticket without taking any further action.

You only see these buttons when the request is in the planned state. Requests that are already in-progress, completed, or cancelled do not show action buttons.

Submitting a New Request

To submit a new feature request or bug report, click New Request above the Requests table. Fill in a title and description, then click Submit. The triage agent picks up the ticket and creates a plan within minutes.

Viewing Pull Requests

When the implementation agent finishes work on a sub-task, it opens a pull request. For apps hosted on GitHub, the PR link points to the GitHub pull request. For Gitea-hosted apps, the link goes to the Gitea pull request. Both appear in the sub-task accordion inside the Request Detail Drawer.

Enable via MCP

If you have an AI agent connected to OSC via MCP, you can enable Agentic SDLC with a natural language request:

Enable Agentic SDLC for my app called myapp

Or use the enable-agentic-sdlc tool directly:

Enable Agentic SDLC for app ID myapp-abc123 using source URL https://github.com/myorg/myapp

The agent accepts an optional gitToken when the repository is private or the app's stored credentials lack write access.

Check Status

What is the Agentic SDLC status for my app called myapp?

The get-agentic-sdlc-status tool returns:

Field Description
status initializing, active, failed, or disabled
initTaskId ID of the bootstrap agent task
scheduleTaskId ID of the recurring scheduled agent task
enabledAt Timestamp when Agentic SDLC was enabled
Bootstrap run status Status of the most recent bootstrap run
Last weekly run status Status of the most recent scheduled run

Disable

To stop autonomous SDLC activity for an app, toggle the switch off in the My Apps detail panel, or use MCP:

Disable Agentic SDLC for my app called myapp

Disabling removes both the bootstrap task and the scheduled task. The CLAUDE.md (or AGENTS.md) that was written to your repository is not removed; it stays in the repo as a useful context file for future manual or automated agent sessions.

Available MCP Tools

Tool Description
enable-agentic-sdlc Enable autonomous SDLC for a My App; accepts appId, optional sourceUrl, gitToken, agentType, and model
disable-agentic-sdlc Disable Agentic SDLC and remove the associated agent tasks
get-agentic-sdlc-status Get current status, task IDs, and last-run results for a given app

Status Reference

Status Meaning
initializing Bootstrap task is running; CLAUDE.md is being written and pushed
active Bootstrap completed; scheduled agent runs are active
failed Bootstrap task failed; check logs in My Agent Tasks
disabled Agentic SDLC is not enabled for this app

Security Audit Add-On

The Agentic SDLC has an optional weekly security audit add-on. When enabled, it runs a set of automated security scanners against your app's source every Monday and converts each new finding into a standard SDLC ticket that flows through the same triage, approve, and implement pipeline.

See Agentic SDLC Security Audit for setup and details.