Configure tools

Tools are the capabilities your users' MCP servers expose to AI agents. The SDK lets you control which tools are available so users get the right context for your product workflow.

Audience	SDK implementers and product teams deciding which AI actions users can take
Prerequisites	A working SDK embed and a product use case for context-backed AI answers

Tool configuration

All 10 tools are enabled by default. In the embed snippet, you only need to specify tools you want to disable:

MCPStudio.init({
  clientId: "YOUR_CLIENT_ID",
  container: "#mcp-studio-widget",
  tools: {
    search_issues: false,   // Disable issue search
    get_changelog: false,   // Disable changelog retrieval
  },
});

Available tools

Search and discovery

Tool	Key	Description
Search Documentation	search_docs	Full-text search across all connected documentation sources
Query Source	query_source	Query a specific source URL for targeted information
Search Issues & Solutions	search_issues	Search for known issues, bugs, workarounds, and solutions

Content extraction

Tool	Key	Description
Get Code Examples	get_code_examples	Extract code snippets and examples from documentation
Extract Schema & Types	extract_schema	Extract data schemas, type definitions, and data models

Analysis

Tool	Key	Description
Summarize Content	summarize_content	Generate a concise summary of a source page or section
Ask a Question	ask_question	Ask a natural language question and get a synthesized answer

Reference and guides

Tool	Key	Description
Find API Reference	find_api_reference	Locate API endpoints, method signatures, and parameters
Get Changelog	get_changelog	Retrieve recent changes, release notes, and version updates
Get Quickstart Guide	get_quickstart	Extract setup instructions and getting-started content

How results are ranked

Servers your end users create through the embed use the same retrieval engine as MCP Studio itself, so the ranking behaviour is identical and you do not configure any of it. Keyword and meaning-based matching run together, a relevance model reorders candidates when the leading scores are close, current documentation is preferred over archived versions, and a response says so when nothing retrieved is a strong match for the question.

That last point matters most when you are embedding. A connected agent will answer confidently from whatever it is handed, so a tidy list of unrelated passages is worse for your users than an admission of uncertainty. Your end users get that disclosure without doing anything, and it is why an MCP server created through your embed is a safer thing to point an agent at than a plain search over the same content.

How it works describes each stage.

Recommended configurations

Documentation-focused

Best for products where users primarily need to search and understand docs:

tools: {
  search_issues: false,
  extract_schema: false,
}

API-focused

Best for API products where schema and endpoint discovery matters most:

tools: {
  get_changelog: false,
  get_quickstart: false,
}

Minimal

Simplified experience with only the essentials:

tools: {
  search_issues: false,
  extract_schema: false,
  get_changelog: false,
  get_quickstart: false,
  summarize_content: false,
  find_api_reference: false,
  query_source: false,
}

info

A minimum of 3 tools must be enabled for the wizard to function properly.