Claude Code + Unmarkdown: Publish Docs Without Leaving the Terminal

Claude Code is Anthropic's CLI for working with Claude directly in your terminal. It reads your codebase, writes code, runs commands, and manages files. With MCP servers, it can also reach into external services.

Connect Unmarkdown™ to Claude Code, and you get document publishing without leaving the terminal. Generate a changelog from your git history, publish it as a styled web page, convert a README for stakeholders, and post a summary to Slack. All from a single conversation in your terminal.

Setup takes 30 seconds.

Setting up Claude Code with Unmarkdown

One command adds the Unmarkdown™ MCP server to Claude Code:

claude mcp add unmarkdown -- npx -y @un-markdown/mcp-server

Claude Code will ask for your API key. Generate one at Settings > API in your Unmarkdown™ account, then paste it when prompted. The key is stored in your Claude Code configuration and used for all subsequent requests.

To verify the connection:

claude mcp list

You should see unmarkdown: connected (stdio) in the output. Claude Code now has access to all 7 Unmarkdown™ tools: create_document, list_documents, get_document, update_document, publish_document, convert_markdown, and get_usage.

Configuration options

User scope (default): The server is available in all your Claude Code sessions. Configuration stored in ~/.claude.json.

claude mcp add --scope user unmarkdown -- npx -y @un-markdown/mcp-server

Project scope: The server is only available when working in a specific project. Configuration stored in .claude.json in the project root, which you can commit to share with your team.

claude mcp add --scope project unmarkdown -- npx -y @un-markdown/mcp-server

Manual configuration

If you prefer to edit the config file directly, add this to your ~/.claude.json:

{
  "mcpServers": {
    "unmarkdown": {
      "command": "npx",
      "args": ["-y", "@un-markdown/mcp-server"],
      "env": {
        "UNMARKDOWN_API_KEY": "um_your_key_here"
      }
    }
  }
}

For background on what MCP servers are and how the protocol works, see What is MCP? A Plain-English Guide for AI Users.

Five workflows you can use today

These are real prompts you can give Claude Code once the connection is set up. Each one uses multiple MCP tools behind the scenes.

1. Generate and publish a changelog

Prompt: "Look at the git log for the last 2 weeks, write a changelog following Keep a Changelog format, and publish it as a web page with the GitHub template."

What happens:

1. Claude reads your git log via Bash
2. Groups commits by type (features, fixes, changes)
3. Calls create_document with the changelog content and github template
4. Calls publish_document to create a shareable URL
5. Returns the published URL in the conversation

Example output:

I've published the changelog at:
https://unmarkdown.com/u/yourname/changelog-feb-2026

The release covers 23 commits across 14 days:
- 5 new features (search, graph, folders, linking, zen mode)
- 8 bug fixes
- 4 refactoring changes
- 6 documentation updates

You get a formatted, published changelog without opening a browser, writing any markdown by hand, or copying anything.

2. Convert README for stakeholders

Prompt: "Read the README.md and convert it for Google Docs. I need to share it with the product team."

What happens:

1. Claude reads README.md from your project
2. Calls convert_markdown with destination: "google-docs" and a professional template
3. Returns the Google Docs-optimized HTML

For stakeholders who need a shareable link instead:

Prompt: "Read the README and publish it as a clean web page with the Executive template."

Claude creates a document, publishes it, and gives you a URL. The published page uses proper heading hierarchy, styled tables, and formatted code blocks, all rendered with the template you chose.

3. Create weekly status reports

Prompt: "Create a weekly status document in my Projects folder. Include: shipped items from git log this week, open issues from GitHub, and a Next Steps section. Use the Consulting template."

What happens:

1. Claude reads git log and GitHub issues via Bash/GitHub CLI
2. Compiles everything into structured markdown
3. Calls create_document with folder: "Projects" and template_id: "consulting"
4. Returns the document ID

Next week:

Prompt: "Update the weekly status document with this week's progress and re-publish it."

Claude calls list_documents to find the existing document, get_document to read it, update_document to add the new week's entry, and publish_document to refresh the published page. The URL stays the same.

This creates a running document that builds up over time, exactly as described in The Document Your AI Never Forgets: Persistent Knowledge with MCP.

4. Batch publish documentation

Prompt: "Find all the markdown files in the docs/ directory, create Unmarkdown documents for each one, and publish them all."

What happens:

1. Claude uses Glob to find docs/**/*.md
2. For each file, reads the content, creates a document with an appropriate title, and publishes it
3. Returns a list of all published URLs

This is useful for quickly making internal documentation accessible to people outside your repository. Each document gets its own URL with proper template styling.

5. Convert meeting notes for Slack

Prompt: "Read the file at notes/standup-2026-02-25.md and convert it for Slack so I can post it in #engineering."

What happens:

1. Claude reads the file
2. Calls convert_markdown with destination: "slack"
3. Returns Slack-formatted text: bold as single asterisks, headings as bold labels, links in <url|text> format, tables as monospace blocks

You copy the output and paste it directly into Slack. The formatting renders correctly because it has been converted to Slack's mrkdwn format rather than pasted as raw markdown.

Chaining with other MCP servers

Claude Code supports multiple MCP servers simultaneously. This lets you build workflows that span services.

GitHub + Unmarkdown: "Look at the open pull requests on this repo, summarize the most important ones, and publish a code review digest."

Brave Search + Unmarkdown: "Research the top 5 competitors in our space, write a competitive analysis, and publish it with the Executive template."

Filesystem + Unmarkdown: "Read all the ADR files in docs/decisions/, compile them into a single architecture decisions document, and publish it."

Each MCP server handles its domain. GitHub provides PR data. Brave Search provides web results. Filesystem provides file access. Unmarkdown™ handles document creation, styling, and publishing. Claude orchestrates the tools based on your request.

To add more servers, use the same claude mcp add command. For example:

claude mcp add github --scope user
claude mcp add brave-search -- npx -y @anthropic/mcp-server-brave-search

Run claude mcp list to verify all servers are connected.

Troubleshooting Claude Code MCP connections

"Server not found" after adding: Restart Claude Code. MCP server changes require a session restart to take effect.

"Authentication failed": Verify your API key is correct. Run claude mcp get unmarkdown to check the configuration. Generate a new key at Settings > API if needed.

"Quota exceeded": Check your usage with the prompt "Check my Unmarkdown usage." Claude calls get_usage and shows your remaining calls. Free tier: 1,000 calls/month. Pro ($8/month annual): 10,000 calls/month.

Slow first call: The first MCP call in a session installs @un-markdown/mcp-server via npx. Subsequent calls are fast. To skip the install delay, install the package globally: npm install -g @un-markdown/mcp-server.

Why terminal-native publishing matters

Developers live in the terminal. Context switching to a browser, logging into a web app, pasting content, clicking through UI elements, and copying a link back kills flow. Every context switch costs attention and time.

Terminal-native publishing keeps you in flow. You are already in Claude Code, already thinking about documentation, already looking at the content that needs to be published. Adding an MCP server turns that thought into a published document without ever leaving the conversation.

For more on using CLAUDE.md files to configure Claude Code's behavior, including documentation workflows, see the dedicated guide. And for the full reference on all 7 MCP tools with detailed examples, see How to Use Claude's MCP Tools to Publish Documents.

Related reading

• How to Use Claude's MCP Tools to Publish Documents
• How to Use CLAUDE.md Files (And When You Need Something More)
• What is MCP? A Plain-English Guide for AI Users
• The Document Your AI Never Forgets: Persistent Knowledge with MCP