adins-prd-mcp

adins-prd-mcp

MCP server for reading Confluence PRD pages and rendering SRS/URS documents as Word files. Built for use with AI agents in the AdIns SDLC toolchain.

Tools

Confluence tools

Database tools

Render tools

---

Workflow

init_document_db(db_path, meta)
  ↓
get_confluence_page_to_file(page_id, output_path)   ← subagent fetches its page to disk
  ↓
store_feature(db_path, feature_data)                ← subagent stores parsed data
  ↓ (up to 5 parallel subagents, one per feature page)
render_srs_docx(template_path, output_path, db_path)
  or
render_urs_docx(template_path, output_path, db_path)

The main agent sets service_order. Each subagent sets feature_order from its sibling position in the Confluence page tree. Menu order is derived from list position. The agent never assembles the full context dict. The render tools read directly from the DB.

---

Requirements

• Python 3.12+
• At least one of the two auth methods configured (see below)

---

Authentication

Two methods are supported. The CONFLUENCE_AUTH_MODE variable controls which one is used. The default (auto) tries Basic first and falls back to OAuth automatically.

Option A — Basic auth (API token)

Generate a classic API token at id.atlassian.com/manage-profile/security/api-tokens.

Required env vars:

Option B — OAuth 2.0 (3LO)

Register an OAuth 2.0 (3LO) app at developer.atlassian.com/console/myapps.

App settings:

• Callback URL: http://localhost:8877/callback
• Scopes required:
  • read:confluence-content.all
  • read:confluence-props
  • read:confluence-space.summary
  • search:confluence
  • readonly:content.attachment:confluence
  • offline_access (enables silent token refresh — check this box)

Required env vars:

First-time login: on the first tool call that needs Confluence access, the MCP server opens your browser to the Atlassian authorisation page. After you approve, the server captures the callback automatically and stores tokens in your OS credential store (Windows Credential Manager / macOS Keychain / libsecret). Subsequent runs reuse the stored tokens and refresh them silently — no browser interaction needed until the refresh token itself expires.

Force re-authentication:

"env": {
  "CONFLUENCE_OAUTH_FORCE_REAUTH": "true"
}

Set this temporarily to clear stored tokens and trigger a fresh browser login. Remove the flag after the new tokens are stored.

---

Configuration

With uvx (recommended) — Basic auth

{
  "mcpServers": {
    "adins-prd-mcp": {
      "command": "uvx",
      "args": ["adins-prd-mcp"],
      "env": {
        "CONFLUENCE_BASE_URL": "https://yourcompany.atlassian.net",
        "CONFLUENCE_EMAIL": "you@yourcompany.com",
        "CONFLUENCE_API_TOKEN": "your-token-here"
      }
    }
  }
}

With uvx — OAuth 2.0

{
  "mcpServers": {
    "adins-prd-mcp": {
      "command": "uvx",
      "args": ["adins-prd-mcp"],
      "env": {
        "CONFLUENCE_BASE_URL": "https://yourcompany.atlassian.net",
        "CONFLUENCE_AUTH_MODE": "oauth",
        "CONFLUENCE_OAUTH_CLIENT_ID": "your-client-id",
        "CONFLUENCE_OAUTH_CLIENT_SECRET": "your-client-secret"
      }
    }
  }
}

With a local install

{
  "mcpServers": {
    "adins-prd-mcp": {
      "command": "python",
      "args": ["-m", "adins_prd_mcp.server"],
      "cwd": "/path/to/mcp",
      "env": {
        "CONFLUENCE_BASE_URL": "https://yourcompany.atlassian.net",
        "CONFLUENCE_EMAIL": "you@yourcompany.com",
        "CONFLUENCE_API_TOKEN": "your-token-here"
      }
    }
  }
}

All environment variables

---

store_feature payload shape

{
  "db_path": "output/LMS-1.5.db",
  "feature_data": {
    "source_id": "935559172",
    "source_version": 24,
    "raw_content_path": "output/LMS-1.5/935559172/raw.md",
    "service_name": "Amendment",
    "service_style": "standard",
    "service_order": 0,
    "feature_name": "Partial Prepayment",
    "feature_description": "Partial Prepayment merupakan fitur ...",
    "feature_order": 2,
    "process_flow_image_path": "output/LMS-1.5/935559172/process-flow.png",
    "menus": [
      {
        "title": "Partial Prepayment Request",
        "actor": "Operation Staff",
        "description": "Menu for submitting a partial prepayment request.",
        "figma_url": "[Figma - Partial Prepayment Request](https://www.figma.com/proto/...)",
        "constraints": [
          {
            "group": "Ada beberapa persyaratan:",
            "children": [
              {"text": "Status agreement tidak sedang dalam proses lain."},
              {"text": "Belum mencapai maksimal amendment limit."}
            ]
          }
        ],
        "boundaries": [
          {"group": "", "children": [{"text": "Fitur ini belum support untuk kontrak syariah."}]}
        ],
        "actions": [
          {"action": "Submit", "description": "Submit the prepayment request for approval."}
        ],
        "scenarios": [
          {"no": "1", "scenario": "Submit with valid data.", "expectation": "Request saved with status Pending."}
        ],
        "gaps": [
          {
            "number": "GAP-001",
            "title": "Prepayment amount validation",
            "summary": "Current system does not validate prepayment amount.",
            "current_system": "No validation exists.",
            "system_solution": "Add server-side validation."
          }
        ]
      }
    ]
  }
}

---

Development

uv sync
uv run adins-prd-mcp

Publishing

uv build
uv publish