esa

This MCP server provides a bridge between an AI assistant and the esa information sharing service. It exposes a set of MCP-compatible commands that let you read, create, update, and manage esa articles, comments, categories, and more, directly from your AI workflows. Tools include team management (esa_get_teams, esa_get_team_stats, esa_get_team_tags, esa_get_team_members), article management (esa_search_posts, esa_get_post, esa_create_post, esa_update_post), article operations (esa_archive_post, esa_ship_post, esa_duplicate_post), comment management (esa_get_comment, esa_create_comment, esa_update_comment, esa_delete_comment, esa_get_post_comments, esa_get_team_comments), category management (esa_get_categories, esa_get_top_categories, esa_get_all_category_paths), attachment access (esa_get_attachment), documentation helpers (esa_get_search_options_help, esa_get_markdown_syntax_help, esa_search_help), and content discovery (esa_recent_posts, esa_summarize_post). To use these in Claude Desktop or any MCP client, configure the mcpServers section with the esa server, provide ESA_ACCESS_TOKEN and LANG, and run the server. The MCP client will then send requests to esa via the available commands and receive structured results suitable for AI processing.

Prerequisites:

• Docker installed and running (recommended for this server).
• An ESA access token with at least read/write permissions (PAT v2 recommended).

Installation steps:

1. Install Docker on your system following the official instructions for your OS.

2. Authenticate or prepare your ESA access token and set your preferred UI language (LANG), typically ja or en.

3. Run the MCP server via Docker using the provided command:

   docker run -i --rm -e ESA_ACCESS_TOKEN=your_personal_access_token -e LANG=ja ghcr.io/esaio/esa-mcp-server

4. In your MCP client configuration (e.g., Claude Desktop config), point to the running docker container by using the docker command approach and ensure the environment variables are passed as shown in the mcp_config example.

5. Verify that the server starts and responds to MCP requests. If you prefer a local development setup, you can alternatively use npx with the same environment variables and the package name if you build it locally, but the README primarily demonstrates the docker approach.

Tips and caveats:

• Ensure ESA_ACCESS_TOKEN includes the required scopes (read/write or admin:comment, etc.).
• LANG controls the UI language in the client; set to ja for Japanese or en for English.
• When using Docker, the --rm flag ensures the container is removed after stopping; remove it if you want to reuse the container.
• If you encounter token or permission errors, re-check the token scopes and re-authenticate.
• For Claude Desktop integration, you can also use the npx option if you prefer running directly from Node.js tooling, but the docker approach is the recommended and most portable setup.
• The MCP server supports a variety of operations; refer to the tool list in the README to discover the full set of available commands.
• If you need to run multiple MCP servers, use distinct server names under mcpServers to avoid conflicts.