Rubber-ducky (Workflow Orchestrator MCP)

Repo: https://github.com/soad666p/Rubber_Ducky MCP server that can be used from any MCP-compatible client (for example Gemini CLI, Claude Desktop, or other LLM tools). It exposes workflow tools (Jira, Confluence, GitHub, Splunk, SonarQube, etc.) and guides through proposals, planning, implementation, testing, log analysis, and PR creation.

Extension Installation (One-Click)

You can install this as a Gemini CLI extension directly from GitHub:

gemini extensions install https://github.com/soad666p/Rubber_Ducky

Note: Local mode now auto-installs dependencies on first launch. You can still run npm install manually in the extension directory (~/.gemini/extensions/rubber-ducky) if you prefer.

Vercel Deployment (Cloud Hosting)

To host your own instance of this MCP server on Vercel:

1. Fork this repository.
2. Deploy to Vercel:
   • Connect your fork to Vercel.
   • Vercel will auto-detect the project as a Node.js project.
3. Configure Environment Variables in Vercel:
   • Add any tokens you need (e.g., JIRA_API_TOKEN, GITHUB_TOKEN).
4. Update gemini-extension.json:
   • In your fork, update the url in gemini-extension.json to point to your Vercel URL (e.g., https://your-app.vercel.app/api/sse).
5. Install:

   gemini extensions install https://github.com/YOUR_USERNAME/Rubber_Ducky

---

MCP client configuration (Manual)

• ~/.gemini/settings.json

Add or edit the mcpServers section there. The gemini-md-mcp entry should point at this repo’s server (see Setup below). A copyable template with placeholders is in gemini-settings.example.json — copy it to your client config and replace placeholders using the token instructions below.

If you are using a client other than Gemini, use the equivalent MCP server configuration location and schema for that client, and point it to this same server entrypoint.

have a look at the PRD

Setup

npm install
npm run build

To run the MCP server directly (e.g. for debugging):

npm run dev

In ~/.gemini/settings.json, ensure gemini-md-mcp points to your clone, for example:

"gemini-md-mcp": {
  "command": "node",
  "args": [
    "--loader",
    "ts-node/esm",
    "/path/to/your/Rubber_Ducky/src/server.ts"
  ],
  "cwd": "/path/to/your/Rubber_Ducky"
}

Use your actual project path instead of /path/to/your/Rubber_Ducky.

---

Getting API tokens for MCPs

The MCPs used by this workflow need API tokens. Configure them under mcpServers in your MCP client config for each server. Below is how to obtain each token and where it goes.

Jira / Atlassian (mcp-atlassian)

• What you need: JIRA_API_TOKEN and optionally CONFLUENCE_API_TOKEN (same token can be used for both).
• How to get it:
  1. Go to Atlassian API tokens.
  2. Click Create API token (or Create API token with scopes if available).
  3. Name it (e.g. “Gemini MCP”), set expiration, choose Jira and/or Confluence, set scopes.
  4. Copy the token and store it securely.
• Where it goes: In mcpServers["mcp-atlassian"].env:
  • JIRA_API_TOKEN
  • CONFLUENCE_API_TOKEN (if using Confluence) Use your Atlassian email for JIRA_USERNAME / CONFLUENCE_USERNAME.

Confluence (confluence-cloud-mcp)

• What you need: CONFLUENCE_API_TOKEN (same as above).
• How to get it: Same as Jira / Atlassian — create an API token at the same Atlassian page and use it for Confluence.
• Where it goes: In mcpServers["confluence"].env:
  • CONFLUENCE_API_TOKEN
  • CONFLUENCE_EMAIL (your Atlassian email)
  • CONFLUENCE_DOMAIN (e.g. yourcompany.jira.com).

Splunk (splunk-mcp-server)

• What you need: A Splunk authentication token (Bearer token) for your Splunk Cloud/Enterprise instance.
• How to get it:
  • Splunk Cloud: Manage authentication tokens (Splunk Cloud docs).
  • Splunk Enterprise: Create authentication tokens. Create a token, then use it as a Bearer token in API requests.
• Where it goes: In mcpServers["splunk-mcp-server"].args, the remote MCP URL is passed to mcp-remote; you need to pass the token, e.g. via --header "Authorization: Bearer YOUR_SPLUNK_TOKEN". Replace YOUR_SPLUNK_TOKEN with the token you created.

GitHub (GitHub Copilot MCP)

• What you need: A GitHub Personal Access Token (PAT) with scopes that allow the operations you need (e.g. repo read, pull requests).
• How to get it:
  1. GitHub → Settings → Developer settings → Personal access tokens (or github.com/settings/tokens).
  2. Generate new token (classic or fine-grained). Grant at least the repo permissions needed for the MCP (e.g. read repo, read/write PRs).
  3. Copy the token once; it won’t be shown again.
• Where it goes: In mcpServers["github"].headers:
  • Authorization: "Bearer ghp_xxxxxxxx" (replace with your PAT).

SonarQube / SonarCloud (sonarqube)

• What you need: A user token from your SonarQube or SonarCloud instance.
• How to get it:
  • SonarCloud: Log in at sonarcloud.io → My Account → Security → Generate Tokens. Name it and copy the token.
  • SonarQube: Log in → My Account → Security → Generate Tokens (or ask an admin to create one for you).
• Where it goes: In mcpServers["sonarqube"].env:
  • SONARQUBE_TOKEN — the token you generated
  • SONARQUBE_URL — e.g. https://sonarcloud.io or your SonarQube URL
  • SONARQUBE_ORG — for SonarCloud, your organization key (e.g. your-org)

---

Workflow

The orchestrator drives a structured flow:

• Proposals: Present multiple implementation options for a change.
• Plan (Conductor-style): After choosing a proposal, produce a spec and phased plan. Artifacts are written under conductor/tracks/<sessionId>/ (e.g. spec.md, plan.md) so the repo can be used with Conductor in Gemini CLI.
• Implementation: Guide code changes.
• Testing: Run and interpret tests.
• Log analysis: Inspect logs for debugging and verification (e2e, etc.).
• PR creation: Help create Pull Requests.

Flow: start → select_proposal → generate_plan → answer questions → implementation → testing → log analysis → PR creation.

Workflow diagram

  flowchart LR
    A[start] --> B[select_proposal]
    B --> C[generate_plan]
    C --> D[answer questions]
    D --> E[implementation]
    E --> F[testing]
    F --> G[log analysis]
    G --> H[PR creation]

Loading

License

See repository license file.