graphql-agent-toolkit
Generates CrewAI tools from GraphQL schemas, enabling AI agents to execute queries and mutations with JSON Schema argument definitions.
Transforms any GraphQL API into AI-agent-ready tools by introspecting the schema, generating operations, and handling pagination and result summarization.
Generates LangChain-compatible tools from GraphQL schemas, supporting both basic and structured tools with Zod schemas.
Supports the Vercel AI SDK by generating tool definitions with Zod schemas and execute functions for seamless integration.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@graphql-agent-toolkitfetch user with ID 123"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
graphql-agent-toolkit
Turn any GraphQL API into AI-agent-ready tools -- MCP servers, LangChain tools, and framework adapters.
graphql-agent-toolkit introspects a GraphQL endpoint, generates typed operations, and exposes them as tools that AI agents can discover and call. It supports the Model Context Protocol (MCP) out of the box, so you can connect any MCP-compatible AI client to any GraphQL API in seconds.
Quick Start
npx graphql-agent-toolkit init --endpoint https://your-api.com/graphqlThis introspects your schema and prints a configuration summary. To start an MCP server:
npx graphql-agent-toolkit serve --endpoint https://your-api.com/graphqlRelated MCP server: graphql-to-mcp
Installation
npm install graphql-agent-toolkit graphqlRequirements
Node.js >= 18.0.0
graphql>= 16.0.0 (peer dependency)TypeScript >= 5.0 (optional, for type definitions)
Fully written in TypeScript with complete type exports for all public APIs.
Features
Schema Introspection -- Automatically fetches and parses any GraphQL schema
Operation Builder -- Generates queries and mutations with proper variable definitions and nested selection sets
MCP Server -- Creates a fully functional MCP server with tools for every query and mutation
Semantic Search -- TF-IDF powered schema navigator to find relevant types and fields
Pagination Handling -- Auto-detects and handles Relay and offset pagination across multiple pages
Result Summarization -- Truncate large responses for LLM context windows with markdown formatting
Framework Adapters -- Generate tools for LangChain, CrewAI, and Vercel AI SDK with zero framework dependencies
Mock Data Generation -- Generate deterministic mock data from your schema with
@mock()directive supportCLI -- Command-line interface for quick setup and serving
Dual Format -- Ships as both ESM and CJS with full TypeScript types
Programmatic API
Introspect and Parse a Schema
import { fetchSchema, parseSchema } from 'graphql-agent-toolkit';
const introspection = await fetchSchema({
endpoint: 'https://your-api.com/graphql',
headers: { Authorization: 'Bearer YOUR_TOKEN' },
});
const schema = parseSchema(introspection);
console.log(`Query type: ${schema.queryType}`);
console.log(`Types: ${schema.types.size}`);Build Operations
import { fetchSchema, parseSchema, buildOperation } from 'graphql-agent-toolkit';
const introspection = await fetchSchema({ endpoint: 'https://your-api.com/graphql' });
const schema = parseSchema(introspection);
const op = buildOperation(schema, 'user', { maxDepth: 3 });
console.log(op.operation);
// query UserQuery($id: ID!) {
// user(id: $id) {
// id
// name
// email
// posts {
// id
// title
// }
// }
// }
console.log(op.variables);
// [{ name: 'id', type: 'ID!', required: true, description: 'User ID' }]Create an MCP Server
import { createAgentToolkitServer } from 'graphql-agent-toolkit';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
const server = await createAgentToolkitServer({
endpoint: 'https://your-api.com/graphql',
headers: { Authorization: 'Bearer YOUR_TOKEN' },
operationDepth: 2,
});
const transport = new StdioServerTransport();
await server.connect(transport);Each query becomes a query_<fieldName> tool, and each mutation becomes a mutate_<fieldName> tool. An additional explore_schema tool lets the agent browse types and fields.
Semantic Schema Navigation
import { fetchSchema, parseSchema, SchemaNavigator } from 'graphql-agent-toolkit';
const introspection = await fetchSchema({ endpoint: 'https://your-api.com/graphql' });
const schema = parseSchema(introspection);
const navigator = new SchemaNavigator();
navigator.index(schema);
// Search for relevant types
const results = navigator.search('user authentication');
for (const result of results) {
console.log(`${result.typeName} (${result.kind}) - score: ${result.score.toFixed(3)}`);
}
// Get detailed context for a type
const context = navigator.getTypeContext('User');
console.log(context);Result Summarization
Truncate large GraphQL responses to fit within LLM context windows:
import { summarizeResponse, formatForLLM } from 'graphql-agent-toolkit';
// Summarize a large response
const { summary, metadata } = summarizeResponse(largeResponse, {
maxItems: 5, // max array items to include
maxDepth: 3, // max nesting depth
maxStringLength: 200, // truncate long strings
includeMetadata: true, // add _meta with counts
});
console.log(metadata);
// { totalItems: 1500, truncated: true, originalSize: 48230 }
// Format as clean markdown for LLM context
const markdown = formatForLLM(largeResponse, { maxItems: 10 });
console.log(markdown);Framework Adapters
Generate tools for popular AI frameworks -- no framework dependencies required.
LangChain
import { createLangChainTools, createStructuredTools } from 'graphql-agent-toolkit';
// Basic tools (input is JSON string)
const tools = createLangChainTools(schema, executor, { maxDepth: 2 });
// Structured tools with Zod schemas (for @langchain/core StructuredTool)
const structuredTools = createStructuredTools(schema, executor);
for (const tool of tools) {
console.log(`${tool.name}: ${tool.description}`);
// tool.func(jsonString) -> Promise<string>
}CrewAI
import { createCrewAITools } from 'graphql-agent-toolkit';
const tools = createCrewAITools(schema, executor);
for (const tool of tools) {
console.log(`${tool.name}: ${tool.description}`);
// tool.args_schema is a JSON Schema object
// tool.func(argsObject) -> Promise<string>
}Vercel AI SDK
import { createVercelAITools } from 'graphql-agent-toolkit';
const tools = createVercelAITools(schema, executor);
// Returns Record<string, { description, parameters: ZodSchema, execute }>
// Use directly with Vercel AI SDK's tool() function
for (const [name, tool] of Object.entries(tools)) {
console.log(`${name}: ${tool.description}`);
// tool.parameters is a Zod schema
// tool.execute(args) -> Promise<string>
}Mock Data Generation
Generate deterministic mock data from your schema for testing:
import { generateMockData, createMockExecutor } from 'graphql-agent-toolkit';
// Generate mock data for a specific type
const mockUser = generateMockData(schema, 'User', {
seed: 42, // deterministic output
arrayLength: 3, // items per list field
maxDepth: 3, // max recursion depth
});
console.log(mockUser);
// { id: 'id_id_0', name: 'mock_name', posts: [...] }
// Create a drop-in mock executor (no HTTP calls)
const mockExecutor = createMockExecutor(schema, { seed: 42 });
// Use it anywhere a GraphQLExecutor is expected
const result = await mockExecutor.execute(
'query { user(id: "1") { id name } }',
{ id: '1' }
);Use the @mock() directive in field descriptions for custom values:
type Product {
"The product name @mock(\"Widget Pro\")"
name: String!
"Current price in USD @mock(29.99)"
price: Float!
"Whether the product is in stock @mock(true)"
inStock: Boolean!
}CLI Usage
init -- Introspect and generate config
graphql-agent-toolkit init \
--endpoint https://your-api.com/graphql \
--header "Authorization: Bearer YOUR_TOKEN" \
--output config.jsonserve -- Start MCP server
# From a config file
graphql-agent-toolkit serve --config config.json
# Directly from an endpoint
graphql-agent-toolkit serve --endpoint https://your-api.com/graphqlMCP Server Usage
Add to your MCP client configuration (e.g., Claude Desktop):
{
"mcpServers": {
"my-graphql-api": {
"command": "npx",
"args": [
"graphql-agent-toolkit",
"serve",
"--endpoint",
"https://your-api.com/graphql"
]
}
}
}Configuration
The AgentToolkitConfig object accepts:
Property | Type | Default | Description |
|
| (required) | GraphQL endpoint URL |
|
|
| HTTP headers for requests |
|
|
| Max depth for generated selection sets |
|
|
| Include deprecated fields |
API Reference
Introspection
fetchSchema(options)-- Fetch introspection query result from a GraphQL endpointparseSchema(introspection)-- Parse raw introspection result into aParsedSchema
Operations
buildOperation(schema, fieldName, options?)-- Generate a GraphQL operation string with variables
MCP
createAgentToolkitServer(config, options?)-- Create a fully configured MCP servercreateToolsFromSchema(schema, executor, options?)-- Create tool definitions from a parsed schemaGraphQLExecutor-- Class for executing GraphQL operations
Semantic
SchemaNavigator-- Class for indexing and searching a GraphQL schema.index(schema)-- Index a parsed schema.search(query, limit?)-- Search for relevant types.getTypeContext(typeName)-- Get formatted context for a type
Pagination
executePaginated(executor, operation, variables, config?)-- Execute a paginated query, collecting all pagesdetectPaginationStyle(schema, typeName)-- Auto-detect Relay or offset pagination from a type
Summarization
summarizeResponse(data, config?)-- Truncate arrays, limit depth, and shorten strings in a responseformatForLLM(data, config?)-- Format data as clean markdown for LLM context
Framework Adapters
createLangChainTools(schema, executor, options?)-- Create LangChain-compatible tools (JSON string input)createStructuredTools(schema, executor, options?)-- Create LangChain StructuredTool-compatible tools (Zod schemas)createCrewAITools(schema, executor, options?)-- Create CrewAI-compatible tools (dict input,args_schema)createVercelAITools(schema, executor, options?)-- Create Vercel AI SDK-compatible tools (Zod parameters, Record)
Mock Data
generateMockData(schema, typeName, config?)-- Generate mock data for a given typecreateMockExecutor(schema, config?)-- Create a mock executor as drop-in replacement for GraphQLExecutor
Types
AgentToolkitConfig-- Configuration objectParsedSchema-- Parsed schema with type mapSchemaType-- Individual type definitionSchemaField-- Field definition with argsGeneratedOperation-- Generated operation with variablesSearchResult-- Semantic search resultSummaryConfig-- Configuration for response summarizationPaginationConfig-- Configuration for paginated queriesMockConfig-- Configuration for mock data generationLangChainToolConfig-- LangChain tool definition shapeCrewAIToolConfig-- CrewAI tool definition shapeVercelAIToolConfig-- Vercel AI SDK tool definition shape
Contributing
Clone the repository
Install dependencies:
npm installRun tests:
npm testBuild:
npm run buildLint:
npm run lint
License
MIT
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/mstuart/graphql-agent-toolkit'
If you have feedback or need assistance with the MCP directory API, please join our Discord server