Cascade MCP

Cascade MCP helps you analyze Figma, Google Docs and Confluence, and write Jira tickets with AI! This open source project provides a collection of MCP (Model Context Protocol) capabilities for Bitovi's Cascading AI Enablement Initiative.

Checkout our introduction video to understand what it can do:

👉 Btw, Bitovi can help you integrate this into your own SDLC workflow: AI for Software Teams

Supported Tools

Supported tools at the https://cascade.bitovi.com/mcp endpoint:

Combined Tools (Multi-provider workflows):

• analyze-feature-scope - Generate scope analysis from Figma designs linked in a Jira epic (identifies features, establishes scope boundaries, surfaces questions before implementation)
• write-shell-stories - Generate shell stories from Figma designs linked in a Jira epic (analyzes screens, downloads assets, creates prioritized user stories using AI)
• write-epics-next-story - Write the next Jira story from shell stories in an epic (validates dependencies, generates full story content, creates Jira issue with acceptance criteria)
• write-story - Generate or refine a Jira story by gathering comprehensive context (parent hierarchy, comments, linked Figma/Confluence/Google Docs) and writing the best possible story
• extract-linked-resources - Fetch a URL (Jira issue, Confluence page, Google Doc) and return its content as markdown with YAML frontmatter listing all discovered linked resources
• review-work-item - Review a Jira work item and generate questions identifying gaps, ambiguities, and missing information (posts review as Jira comment)

Atlassian/Jira Tools:

• atlassian-get-sites - Get list of accessible Atlassian sites for the authenticated user
• atlassian-get-issue - Retrieve complete details of a Jira issue by ID or key, including description, attachments, comments, and full field data
• atlassian-get-attachments - Fetch Jira attachments by attachment ID (returns base64-encoded content)
• atlassian-update-issue-description - Update a Jira issue description with markdown content (automatically converted to ADF)
• atlassian-add-comment - Post a comment to a Jira issue (markdown automatically converted to ADF)
• atlassian-update-comment - Update an existing Jira comment (markdown automatically converted to ADF)
• confluence-analyze-page - Fetch a Confluence page, convert to markdown, score relevance with LLM, and cache the result
• search - Search Jira issues using JQL (Jira Query Language) with standardized document format output
• fetch - Fetch Jira issue details by issue key/ID in standardized document format (ChatGPT-compatible)

Figma Tools:

• figma-get-user - Get information about the authenticated Figma user (test tool for OAuth validation)
• figma-review-design - Analyze Figma screen designs and post clarifying questions as comments on relevant frames
• figma-batch-zip - Batch-fetch Figma data for multiple URLs; returns a zip download containing per-frame images, structure XML, comments, notes, and analysis prompts (was figma-batch-load)
• figma-batch-cache - Batch-fetch Figma data into server-side cache; returns a batchToken for retrieval via figma-frame-data (cloud environments)
• figma-frame-data - Retrieve a single frame's data (image, context, XML) from batch cache or live Figma
• figma-ask-scope-questions-for-page - [DEPRECATED] Analyze a Figma page and generate design scope questions with frame data and embedded analysis prompts
• figma-post-comment - Post a comment to a Figma file, optionally pinned to a specific frame node
• figma-get-comments - Read existing comments from a Figma file, grouped into threads
• figma-get-image-download - Download images from Figma design URLs (returns base64-encoded image and metadata)
• figma-get-metadata-for-layer - Get detailed metadata for a specific Figma layer including positioning and visual properties
• figma-get-layers-for-page - List all top-level layers from a Figma page with layer IDs, names, types, and download URLs

Google Tools:

• drive-about-user - Retrieve information about the authenticated Google Drive user (test tool for OAuth validation)
• drive-doc-to-markdown - Convert a Google Docs document to Markdown format (supports headings, formatting, lists, and hyperlinks)
• sheets-list-spreadsheets - List Google Spreadsheets accessible to the authenticated user, with optional name filtering
• sheets-get-info - Get spreadsheet metadata including title, locale, and details about each sheet tab
• sheets-read-values - Read cell values from a range in a Google Spreadsheet using A1 notation
• sheets-write-values - Write values to or clear a range in a Google Spreadsheet

Utility Tools:

• utility-test-sampling - Test sampling functionality by sending prompts to the agent and logging the interaction (enables testing of agent capabilities and inter-MCP tool communication)

Read the REST API documentation for accessing these endpoints:

• POST /api/analyze-feature-scope
• POST /api/write-shell-stories
• POST /api/write-next-story

LLM Client Support

The API supports 8 major LLM clients (Anthropic, OpenAI, Google, AWS Bedrock, Mistral, DeepSeek, Groq, xAI). Users choose their LLM client and supply credentials via request headers.

See the LLM Provider Guide for complete documentation.

Use

There are four main ways to use CascadeMCP:

• With a skill-supporting coding agent like Claude or Copilot (see Agent Getting Started)
• With the mini MCP client hosted at https://cascade.bitovi.com/
• With an MCP client that has sampling capabilities (like VSCode Copilot), shown below.
• With a Jira automation, shown here.

If you're just trying to see it work, we recommend using the mini MCP client.

Prerequisites

If you want to do the story writing workflow, you'll need:

• Figma design(s) for some functionality that will need stories
• Access to a Jira instance. Here's a video showing how to create a new Jira instance
• A Jira epic with links to the figma designs and any additional context about what's in and out of scope for the epic

Mini MCP Client

In this example, you'll use the mini MCP client hosted at https://cascade.bitovi.com/ to create stories from a Jira epic.

In order to use the mini MCP client, you'll need an Anthropic SDK API token.

1. Go to https://cascade.bitovi.com/
2. Enter your Anthropic SDK API (this isn't sent to the server, but if anyone asks you to share a key like this, make sure you give one you can delete after).
3. Click connect
4. Authorize Jira and Figma

VSCode Copilot

In this example, we will use VSCode to connect to CascadeMCP and ask CascadeMCP to analyze some Figma images, write shell stories, and the write a story.

Before using the tool, you'll need the following:

• Visual Studio Code with Github Copilot Download here

In your project, create a .vscode/mcp.json file

Click the Add Server... button:

Select the HTTP option:

Paste the following URL and hit enter: https://cascade.bitovi.com/mcp. Then, add cascade-mcp as the name and hit enter.

You can use the following JSON too if the Add Server button did not work for you:

{
    "servers": {
        "cascade-mcp": {
            "url": "https://cascade.bitovi.com/mcp"
        }
    }
}

This will kick off the authentication:

When complete, you should be able to set your copilot chat in Agent mode and ask the question:

MCP: what tools do I have available?

This should tell you you have the Jira MCP tool available: