gitbook-mcp
Read-only MCP server for accessing GitBook content, including spaces, pages, and search, through AI assistants.
README
GitBook MCP Server (Looqbox fork, read-only)
A Model Context Protocol (MCP) server that provides access to GitBook's API for AI assistants and LLM applications.
This is a Looqbox fork of rickysullivan/gitbook-mcp, trimmed to the read tools relevant to docs-editing workflows, with a hardened API client (429 retry, typed error messages). See DESIGN.md for the full design history.
Read-only. Write tools (create_page, update_page) were designed but turned out to be unbuildable against GitBook's live API — see DESIGN.md §9 for findings. Revisit if GitBook ships real write support.
Overview
The GitBook MCP server enables read access to GitBook Organizations, Spaces, and Content through a standardized MCP interface. It provides 7 tools for content retrieval and 6 AI-powered prompts for documentation workflows.
Quick Setup
Prerequisites
- Node.js 20+
- GitBook API token, per-dev personal access token (obtain from https://app.gitbook.com/account/developer)
- Your GitBook Organization ID (optional but recommended)
This fork isn't published to npm — clone and build it locally first:
git clone https://github.com/renancavalcantercb/gitbook-mcp.git
cd gitbook-mcp
npm install
npm run build
Every config below points at the built dist/index.js with node — substitute your actual clone path.
IDE and AI Assistant Integration
Claude Code
Recommended: register per-dev via the CLI, default scope is local (your personal settings, never committed — each dev mints their own PAT):
claude mcp add gitbook-mcp \
-e GITBOOK_API_TOKEN=gb_api_your_token_here \
-e GITBOOK_SPACE_ID=your_default_space_id_here \
-- node /absolute/path/to/gitbook-mcp/dist/index.js --organization-id=your_organization_id_here
If you want a project-shared template instead, drop this in .mcp.json at the repo root with placeholders only — never commit a real token:
{
"mcpServers": {
"gitbook-mcp": {
"type": "stdio",
"command": "node",
"args": [
"/absolute/path/to/gitbook-mcp/dist/index.js",
"--organization-id=your_organization_id_here"
],
"env": {
"GITBOOK_API_TOKEN": "${GITBOOK_API_TOKEN}",
"GITBOOK_SPACE_ID": "your_default_space_id_here"
}
}
}
}
${GITBOOK_API_TOKEN} expands from each dev's own shell environment at runtime — no token ever lands in the committed file.
VS Code (with GitHub Copilot)
Add to your VS Code MCP settings:
{
"servers": {
"gitbook-mcp": {
"type": "stdio",
"command": "node",
"args": [
"/absolute/path/to/gitbook-mcp/dist/index.js",
"--organization-id=your_organization_id_here"
],
"env": {
"GITBOOK_API_TOKEN": "gb_api_your_token_here"
}
}
}
}
Claude Desktop
Add to your Claude Desktop configuration (%APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"gitbook-mcp": {
"command": "node",
"args": [
"/absolute/path/to/gitbook-mcp/dist/index.js",
"--organization-id=your_organization_id"
],
"env": {
"GITBOOK_API_TOKEN": "gb_api_your_token_here"
}
}
}
}
See https://modelcontextprotocol.io/quickstart/user for details.
GitHub Copilot plugin for JetBrains IDEs (IntelliJ IDEA, WebStorm, etc.):
Add to your GitHub Copilot MCP settings for JetBrains IDEs (the path may vary by product and OS, e.g., ~/.config/github-copilot/intellij/mcp.json for IntelliJ on Linux/macOS, or the equivalent directory for your JetBrains IDE and platform):
{
"servers": {
"gitbook-mcp": {
"command": "node",
"args": [
"/absolute/path/to/gitbook-mcp/dist/index.js",
"--organization-id=your_organization_id_here"
],
"env": {
"GITBOOK_API_TOKEN": "gb_api_your_token_here"
}
}
}
}
JetBrains AI Assistant
Add to your JetBrains AI Assistant MCP configuration (see official docs for the exact path):
Getting Your GitBook Credentials
- API Token: Visit https://app.gitbook.com/account/developer to generate your API token
- Organization ID: Use the
list_organizationstool after setup to find your organization ID - Space ID (optional): Use the
list_spacestool (requires a valid organization ID and API token) to find specific space IDs
Configuration Options
You can configure the server using:
- CLI arguments:
--organization-id,--space-id - Environment variables in MCP config: Set via the
envobject in your MCP configuration - System environment variables:
GITBOOK_API_TOKEN,GITBOOK_ORGANIZATION_ID,GITBOOK_SPACE_ID,GITBOOK_API_BASE_URL
Note: .env/.env.local files are loaded automatically when running the server from a local build (node dist/index.js) — see Installation & Setup below.
API Reference
Tools
The GitBook MCP server provides 12 tools organized into functional categories. Each tool includes behavioral hints:
- 📖 Read-only: Tool only reads data and doesn't modify anything
- 🔄 Idempotent: Repeated calls with same args have no additional effect with the same result
- 🌐 Open-world: Tool interacts with external entities
Organization Discovery
List Organizations (list_organizations) 📖 🔄 🌐
Lists all accessible GitBook organizations.
Parameters: None
Returns:
{
"organizations": [
{
"id": "string",
"title": "string",
"urls": {
"app": "string",
"public": "string"
}
}
]
}
Space Management
List Spaces (list_spaces) 📖 🔄 🌐
Lists spaces, optionally filtered by organization.
Parameters:
organizationId(optional): Organization ID to filter spaces
Returns:
{
"spaces": [
{
"id": "string",
"title": "string",
"visibility": "string",
"urls": {
"app": "string",
"public": "string"
}
}
]
}
Get Space Details (get_space) 📖 🔄 🌐
Retrieves detailed information about a specific space.
Parameters:
spaceId(required): The ID of the space to retrieve
Returns:
{
"id": "string",
"title": "string",
"description": "string",
"visibility": "string",
"urls": {
"app": "string",
"public": "string"
}
}
Get Space Content (get_space_content) 📖 🔄 🌐
Retrieves the content structure and pages of a space.
Parameters:
spaceId(optional): The ID of the space (uses default if configured)
Returns:
{
"pages": [
{
"id": "string",
"title": "string",
"slug": "string",
"path": "string"
}
]
}
Search Content (search_content) 📖 🔄 🌐
Searches for content within a space using full-text search.
Parameters:
query(required): Search query stringspaceId(optional): The ID of the space to search (uses default if configured)
Returns:
{
"results": [
{
"id": "string",
"title": "string",
"excerpt": "string",
"url": "string"
}
]
}
Content Retrieval
Get Page Content (get_page_content) 📖 🔄 🌐
Retrieves the content of a specific page.
Parameters:
pageId(required): The ID of the page to retrievespaceId(optional): The ID of the space containing the pageformat(optional): Output format ("document"or"markdown", defaults to"document")metadata(optional): Include revision metadata (boolean, defaults tofalse)computed(optional): Include computed revision data (boolean, defaults tofalse)
Returns:
{
"id": "string",
"title": "string",
"content": "string",
"format": "string"
}
Get Page by Path (get_page_by_path) 📖 🔄 🌐
Retrieves page content using the page path.
Parameters:
pagePath(required): The path of the page to retrievespaceId(optional): The ID of the space containing the page
Returns:
{
"id": "string",
"title": "string",
"content": "string",
"path": "string"
}
Dropped from the upstream base (out of scope for docs-editing use case, see
DESIGN.md§3/§8):get_space_files,get_file,list_collections,get_collection,get_collection_spaces.
Prompts
The GitBook MCP server provides 6 AI-powered prompts for documentation workflows:
Fetch Documentation (fetch_documentation)
Fetches and analyzes GitBook documentation content for specific topics.
Parameters:
topic(required): The topic or subject to search for and analyzespaceId(optional): The ID of the space to search (uses default if configured)includeStructure(optional): Set to "true" to include space structure
Returns: A comprehensive analysis of documentation related to the specified topic, including:
- Relevant pages and sections
- Content summaries
- Gaps or areas needing improvement
Analyze Content Gaps (analyze_content_gaps)
Identifies gaps and missing content in documentation.
Parameters:
spaceId(optional): The ID of the space to analyze (uses default if configured)comparisonSource(optional): Source to compare against (default: "internal analysis")
Returns: A detailed gap analysis including:
- Missing topics and incomplete sections
- Coverage gaps prioritized by importance
- Suggestions for new content areas
Content Audit (content_audit)
Performs quality audits of documentation content.
Parameters:
spaceId(optional): The ID of the space to audit (uses default if configured)auditCriteria(optional): Specific criteria to audit (default: "general quality and consistency")
Returns: A comprehensive quality assessment including:
- Content quality and consistency review
- Outdated information identification
- Writing style and formatting recommendations
Documentation Summary (documentation_summary)
Generates comprehensive summaries of GitBook spaces.
Parameters:
spaceId(optional): The ID of the space to summarize (uses default if configured)summaryType(optional): Type of summary - "overview", "technical", "user-guide", or "custom" (default: "overview")
Returns: A structured summary including:
- Space structure and content organization
- Main topics and themes
- Target audience and use cases
Content Optimization (content_optimization)
Optimizes content for SEO, readability, structure, or performance.
Parameters:
spaceId(optional): The ID of the space to optimize (uses default if configured)optimizationType(required): Type of optimization - "SEO", "readability", "structure", or "performance"targetMetrics(optional): Specific metrics or goals to optimize for
Returns: Optimization recommendations including:
- Specific improvement strategies
- Priority-ranked optimization opportunities
- Implementation guidance
Propose Page Edit (propose_page_edit)
Drafts a create/update proposal for a GitBook page for human review. There is no write tool in this server — this prompt never calls the GitBook API to write anything; it only produces markdown for you to paste into the GitBook editor yourself. See DESIGN.md §9–10 for why direct writes aren't possible against the live API.
Parameters:
spaceId(optional): The ID of the space to target (uses default if configured)pagePath(optional): Path of the existing page to update, or intended path for a new page. Omit if undecided.changeDescription(required): What should change, e.g. "VPN IPs changed, update with new values"
Returns: A structured edit proposal including:
- Target title, path, and GitBook app URL
- Mode (update or create)
- Full markdown draft
- Change summary (update) or placement rationale (create)
- An explicit "Not applied — paste this into GitBook yourself" closing line
Refuses outright for any space ID listed in GITBOOK_EXCLUDED_SPACE_IDS (defaults to the public docs space) — see Environment Variables below.
Configuration Reference
The GitBook MCP server supports multiple configuration methods with the following precedence (highest to lowest):
- CLI Arguments - Passed when starting the MCP server
- Configuration Files - Embedded in project configuration files
- Environment Variables - Set in
.env.localor system environment
Environment Variables
| Variable | Required | Type | Description |
|---|---|---|---|
GITBOOK_API_TOKEN |
Yes | string | GitBook API token (obtain from https://app.gitbook.com/account/developer) |
GITBOOK_ORGANIZATION_ID |
No | string | Default organization ID for operations |
GITBOOK_SPACE_ID |
No | string | Default space ID for single-space projects |
GITBOOK_API_BASE_URL |
No | string | API base URL, default https://api.gitbook.com/v1 |
GITBOOK_EXCLUDED_SPACE_IDS |
No | string | Comma-separated space IDs propose_page_edit refuses to draft for, default 4enQ5nGf8OInfkQDrZoR |
Note: Environment variables can be set in
.env.local,.env, or your system environment.
CLI Arguments
| Argument | Alias | Type | Description |
|---|---|---|---|
--organization-id |
--org |
string | Organization ID to work with |
--space-id |
--space |
string | Default space for operations |
Example:
node dist/index.js --organization-id your-org-id --space-id your-space-id
Additional Configuration Files
Typically these files are provided as context to the AI assistant, which means you can store project-based configuration.
.github/copilot-instructions.md.cursorrules.cursor/rules/rules.md.cursor/rules/instructions.md
e.g.
Format:
## GitBook Configuration
For GitBook MCP operations, use the following configuration:
- organization-id: your-org-id-here
- space-id: your-space-id-here
Default Parameter Behavior
When GITBOOK_ORGANIZATION_ID or GITBOOK_SPACE_ID are configured:
- Tools marked as "optional" can omit the corresponding ID parameters
- The configured default values will be used automatically
- Explicit parameters in tool calls override defaults
Development
Prerequisites
- Node.js 20+
- npm
- GitBook API token (obtain from https://app.gitbook.com/account/developer)
Installation & Setup
git clone https://github.com/renancavalcantercb/gitbook-mcp.git
cd gitbook-mcp
npm install
npm run setup
# Add your GITBOOK_API_TOKEN (and optionally GITBOOK_ORGANIZATION_ID, GITBOOK_SPACE_ID) to .env.local
npm run build
Development
npm run dev
Debugging
DEBUG=1 npm run dev
Add the MCP to VS Code for development
You will need to use node as the command when running locally.
The first arg should be the path to the compiled JavaScript output (e.g., dist/index.js).
{
"servers": {
"gitbook-mcp-dev": {
"type": "stdio",
"command": "node",
"args": [
"/my/path/to/gitbook-mcp/dist/index.js",
"--organization-id=your_organization_id_here"
],
"env": {
"GITBOOK_API_TOKEN": "gb_api_your_token_here"
}
}
}
}
Testing
There are currently no unit or integration tests; running npm run test only checks that the TypeScript code compiles successfully (type-check/build verification), and does not execute any actual tests.
npm run test
Error Handling
The API client retries 429 up to twice with exponential backoff before giving up, and maps common statuses to a clear message instead of a raw HTTP error:
| Error Code | Tool error message returned |
|---|---|
401 |
"Invalid or expired token — regenerate PAT in GitBook account settings." |
403 |
"Token's user has no access to this space." |
404 |
"Space or page not found at that path." |
429 |
"Rate limited, try again shortly." (after 2 retries) |
| other | Raw GitBook API error message, passed through |
Troubleshooting
Token Issues:
- Ensure token starts with
gb_api_ - Verify token has not expired
- Check token permissions in GitBook settings
ID Resolution:
- Use
list_organizationsto find valid organization IDs - Use
list_spacesto find valid space IDs - Use
get_space_contentto find valid page IDs
Configuration Issues:
- Verify environment variables are properly set
- Check file permissions on configuration files
- Ensure CLI arguments are properly formatted
Contributing
This is a Looqbox-internal fork of rickysullivan/gitbook-mcp. For changes specific to this fork (read-only trim, error mapping, write-capability findings), open a PR against renancavalcantercb/gitbook-mcp. For upstream fixes unrelated to the fork's changes, consider contributing to the original repo instead.
License
MIT License - see the LICENSE file for details.
Related Documentation
DESIGN.md— design history, including the write-capability blocker (§9)- GitBook API Documentation
- Model Context Protocol Specification
- Detailed Prompt Documentation
Disclaimer
This project is independently developed and is not officially affiliated with, endorsed by, or sponsored by GitBook.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。