AI Delivery MCP

AI Delivery MCP exposes tools for automating task delivery:

• run_task_delivery
• create_merge_request
• update_jira
• generate_release_note

The recommended first path is run_task_delivery in dry-run mode. It validates the workflow shape without creating GitLab or Jira side effects.

Setup

npm install
cp .env.example .env
npm test
npm run build

Environment

See .env.example for GitLab, Jira, release note, and evidence settings.

Dry-Run Verification

npm run build
DELIVERY_DRY_RUN=true npm run smoke:dry-run

Expected output includes:

{
  "status": "dry_run",
  "evidencePath": "artifacts/delivery-evidence/<timestamp>-DRY-123.json"
}

The evidence JSON includes the merge request preview payload, Jira dry-run result, release note markdown, and step-by-step execution status.

MCP Usage

Build the server:

npm run build

Run with stdio:

node dist/src/index.js

Configure your MCP client to launch node /absolute/path/to/dist/src/index.js with the needed environment variables.

Real Integration Verification

Use a test GitLab project and a test Jira issue first. Run run_task_delivery with dryRun: false, then verify:

• GitLab MR title, description, reviewers, labels, or metadata.
• Jira comment or transition.
• Release note content and evidence JSON.

Definition of Done Evidence

For dry-run verification:

• MR preview payload is returned by run_task_delivery.
• Jira comment preview is included in the delivery result.
• Release note markdown is generated from branch and ticket context.
• Evidence JSON is written under artifacts/delivery-evidence.

For real verification:

• Use a test GitLab branch and test Jira issue.
• Run run_task_delivery with dryRun: false.
• Capture the GitLab MR URL, Jira comment or transition evidence, and generated release note path.