JSON Input
Schema Output
jsontoschema
140 tokens56 tokens60% smaller
How to convert JSON to a schema
- Paste your JSON into the left panel, or drag and drop a
.jsonfile. The editor accepts any valid JSON: objects, arrays, or nested structures. - Select your output format using the dropdown: compact schema for LLM prompts, or TypeScript interfaces for type-safe code.
- Review the output in the right panel. The conversion runs instantly as you type. Toggle the Examples switch to include sample values in schema output.
- Copy the result to your clipboard with one click. Use it in your codebase, LLM prompt, API documentation, or validation layer.
Advanced type inference
The conversion engine analyzes every value in your JSON to produce accurate types. Strings, numbers, booleans, and null values are detected automatically. Nested objects generate their own type definitions, keeping the output clean and readable.
Arrays receive special treatment. When array items share the same shape, the engine outputs a single typed array. When items differ, it merges them into one structure with optional fields, so nothing is lost. Mixed-type arrays produce union types in TypeScript mode (e.g. string | number).
Nullable fields
Fields with null values are typed as unions (e.g. string | null), preserving API semantics.
Optional properties
Properties that appear in some array items but not others are marked with ? in TypeScript output.
Deep nesting
Objects nested at any depth are recursively analyzed and output as distinct, named interfaces.
Reduce LLM token usage by 90%+
Including raw JSON in LLM prompts wastes context window tokens. A 200-item API response repeats the same field names hundreds of times. The model needs to see each field name once to understand the structure.
A compact schema captures the full shape of your data (field names, types, nesting, optionality) in a fraction of the tokens. The built-in token counter shows the exact reduction. Typical savings range from 85% to 99% depending on the payload size.
This approach works with any LLM: GPT-4o, Claude, Gemini, Llama, Mistral. Fewer tokens means lower cost per request, more room for instructions, and faster inference.
Generate TypeScript interfaces from JSON
Working with a third-party API that has no official type definitions? Paste a sample response and get clean TypeScript interfaces instantly. The output uses export interface declarations with proper naming, optional markers, and union types where needed.
The generated interfaces are compatible with strict: true in your tsconfig and work with libraries like React, tRPC, React Hook Form, and Zod. Copy them directly into your codebase or use them as a starting point for more specific type definitions.
100% client-side processing
Every conversion runs locally in your browser. The JSON you paste is never transmitted over the network, stored on a server, or logged anywhere. There are no API calls, no analytics on your input data, and no third-party scripts with access to the editor content.
This makes the tool safe for proprietary API responses, internal data structures, and any JSON you would not want to share with a third-party service.
Available conversions
jsontoschema supports multiple output formats from a single JSON input. Use the tool above or visit a dedicated page for each conversion type.
Built for real developer workflows
LLM prompt engineering
Replace raw JSON payloads in your prompts with compact schemas. Preserve structural context while freeing up tokens for instructions.
API documentation
Generate readable type definitions from real API responses. Paste them into your docs, README, or OpenAPI spec as a starting point.
Frontend type safety
Quickly type untyped API endpoints in your TypeScript frontend. No need to manually write interfaces for every response shape.
Runtime validation
Use the generated schema as a reference to build Zod or JSON Schema validators. Bridge the gap between compile-time types and runtime data.
Frequently asked questions
How does the JSON to schema conversion work?
The converter parses your JSON input directly in the browser, analyzes the structure, infers types for every field (including nested objects and mixed arrays), and outputs a compact schema notation. No data is sent to any server.
What is the difference between a compact schema and JSON Schema?
JSON Schema (Draft 7 or 2020-12) is a full specification with verbose syntax designed for validation engines. A compact schema strips away the boilerplate and keeps only what matters: field names, types, and structure. This makes it ideal for LLM prompts where every token counts.
Can I generate TypeScript interfaces from JSON?
Yes. Switch to TypeScript mode using the dropdown in the header. The converter infers types from your JSON values, handles optional fields that appear in some array items but not others, and generates clean export interface declarations.
Is my data safe when using this tool?
All processing happens client-side in your browser using JavaScript. Your JSON never leaves your machine. There are no network requests, no server-side processing, and no data storage.
Why use a schema instead of raw JSON in LLM prompts?
Raw JSON includes repeated values across array items, wasting tokens on data the model does not need. A schema captures the same structural information (field names, types, nesting) in a fraction of the tokens, often reducing usage by 90% or more.
Does the converter handle complex nested JSON?
Yes. The type inference engine recursively analyzes nested objects, arrays of objects, arrays of mixed types (generating union types), and nullable fields. Optional properties that appear in some array items but not others are correctly marked.
Learn more
Read technical guides on JSON schemas, TypeScript type generation, LLM prompt optimization, and developer tooling on the jsontoschema blog.