Providers

Providers connect AI-DLC to external systems your team already uses — ticketing, specs, design tools, and communication channels. When configured, AI-DLC automatically syncs work across these systems during elaboration and execution.

Provider Categories

Category	Purpose	Supported Types
ticketing	Track work as tickets/issues linked to units	Jira, Linear, GitHub Issues, GitLab Issues
spec	Pull requirements and acceptance criteria	Notion, Confluence, Google Docs
design	Reference designs and component specs	Figma
comms	Post status updates and notifications	Slack, Teams, Discord
VCS hosting	Auto-detected from git remote	GitHub, GitLab, Bitbucket
CI/CD	Auto-detected from repo config files	GitHub Actions, GitLab CI, Jenkins, CircleCI

VCS hosting and CI/CD are detected automatically — you don't need to configure them.

Configuration

Add a providers section to your .ai-dlc/settings.yml:

providers:
  ticketing:
    type: jira
    config:
      project_key: "PROJ"
    instructions: |
      - Set ticket type to "Feature"
      - Use Fibonacci story points (1, 2, 3, 5, 8, 13)
      - Map unit discipline to labels: backend → "Elixir", frontend → "ReactNative"

  spec:
    type: notion
    config:
      workspace_id: "your-workspace-id"
    instructions: |
      - Link each ticket to the specific AC page it covers

  design:
    type: figma
    instructions: |
      - Only reference designs marked "Ready for Dev"

  comms:
    type: slack
    config:
      channel: "#dev-updates"

Each provider entry has three fields:

• type (required) — Which tool to connect to
• config (optional) — Provider-specific settings like project keys or workspace IDs
• instructions (optional) — Project-specific rules that customize default behavior

How It Works

During Elaboration

When you run /elaborate, AI-DLC uses providers to enrich the process:

1. Spec provider — Searches for existing requirements, PRDs, and design docs related to your intent
2. Design provider — Pulls relevant mockups and component specs
3. Ticketing provider — After writing artifacts, creates an epic and tickets:
   • One epic per intent (or links to an existing epic if provided by product)
   • One ticket per unit, linked to the epic
   • Unit depends_on relationships become ticket blocked-by links
   • Epic and ticket keys are stored in intent/unit frontmatter

During Execution

Hats interact with providers as they work:

• Builder — Updates ticket to "In Progress" when starting a unit, "Done" on completion, "Blocked" if stuck
• Reviewer — Posts review outcomes as ticket comments. Updates ticket status on approval/rejection
• Comms — Posts notifications when elaboration completes, reviews finish, or blocking issues arise

Graceful Degradation

Provider interactions are advisory (SHOULD, not MUST). If MCP tools aren't available for a configured provider, AI-DLC skips the integration silently and continues working. Missing providers never block your workflow.

Three-Tier Instruction Merge

Provider instructions are merged from three sources, with later tiers supplementing earlier ones:

1. Built-in defaults — Ship with the plugin. Cover universal behaviors like DAG-to-blocked-by mapping for ticketing.
2. Inline instructions — The instructions: field in your settings.yml. Project-specific rules.
3. Project overrides — Markdown files at .ai-dlc/providers/{type}.md for detailed conventions.

Example: Project-Level Override

Create .ai-dlc/providers/jira.md for detailed Jira conventions:

---
provider: jira
type: ticketing
---

# Jira Conventions

## Required Fields
- Story points (Fibonacci: 1, 2, 3, 5, 8, 13)
- Component: Backend | Frontend | Infrastructure
- Sprint: Current sprint unless explicitly backlogged

## Naming Conventions
- Epic titles: "{Quarter} - {Objective}"
- Story titles: Imperative verb (e.g., "Add JWT validation middleware")

Existing Epic Support

If your product team creates epics ahead of time, set the epic field in your intent frontmatter before elaborating:

---
epic: "PROJ-123"
---

AI-DLC will link all tickets to that existing epic instead of creating a new one. If epic is left empty, a new epic is created automatically.

MCP Requirements

Providers work through MCP (Model Context Protocol) tool servers. For each provider type, you need the corresponding MCP server configured in your Claude Code settings:

Provider Type	MCP Tool Pattern
Jira	mcp__*jira*
Linear	mcp__*linear*
GitHub Issues	gh issue (built-in)
Notion	mcp__*notion*
Confluence	mcp__*confluence*
Figma	mcp__*figma*
Slack	mcp__*slack*

Next Steps

• Quick Start — Get started with AI-DLC
• Workflows — Understand the hat-based workflow system
• Cowork Mode — Work on repos remotely without local checkout