Extract JSON from LLM API

Extract valid JSON from messy LLM outputs via REST API. LLMs frequently wrap JSON in markdown code fences, add explanatory text before and after, use single quotes, or include trailing commas — this API strips all of that and returns clean, parseable JSON.

How it works

POST the raw LLM output string. The API tries multiple extraction strategies in order: strip markdown code fences, extract the largest JSON-like substring, fix common syntax errors (single quotes, trailing commas, missing braces), and return valid JSON.

Use cases

• Parse structured data from LLM responses in agent pipelines reliably
• Handle inconsistent JSON formatting across different model providers
• Build robust agent tool call parsers that handle malformed outputs
• Pre-process LLM outputs before passing to JSON Schema validation

API Reference

POST https://slopshop.gg/v1/llm-output-extract-json
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

Input parameters

Example response

{
  "data": {
    "result": "success",
    "processed": true
  },
  "meta": {
    "credits_used": 1,
    "engine": "real",
    "ms": 4
  }
}

Examples

Three real-world scenarios showing how developers use Extract JSON from LLM in production.

curl -X POST https://slopshop.gg/v1/llm-output-extract-json \
  -H "Authorization: Bearer $SLOPSHOP_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Here is the structured data:\n```json\n{\"name\": \"Alice\", \"score\": 92}\n```\nLet me know if you need anything else."}'

curl -X POST https://slopshop.gg/v1/llm-output-extract-json \
  -H "Authorization: Bearer $SLOPSHOP_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "{'name': 'Bob', 'active': true, 'score': 88}"}'

curl -X POST https://slopshop.gg/v1/llm-output-extract-json \
  -H "Authorization: Bearer $SLOPSHOP_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Based on my analysis, the sentiment data is: {\"sentiment\": \"positive\", \"confidence\": 0.94, \"aspects\": [\"battery\", \"design\"]}. These results indicate a strong positive response."}'

Code examples

curl

curl -X POST https://slopshop.gg/v1/llm-output-extract-json \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "your data here"}'

Python

import requests

response = requests.post(
    "https://slopshop.gg/v1/llm-output-extract-json",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"input": "your data here"}
)
result = response.json()
print(result["data"])

Node.js

const response = await fetch("https://slopshop.gg/v1/llm-output-extract-json", {
  method: "POST",
  headers: {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({ input: "your data here" })
});
const { data } = await response.json();
console.log(data);

CLI

# Install the Slopshop CLI
npm install -g slopshop

# Set your API key
export SLOPSHOP_KEY=your_api_key

# Call llm-output-extract-json
slop llm-output-extract-json '{"input": "your data here"}'

Pricing

Credits are purchased in bundles starting at $1 for 1,000 credits. All compute APIs like this one use 1 credit per call — that's $0.001. See all pricing tiers.

Related APIs in Agent Tools

• Validate LLM Output API — Validate LLM output against a JSON schema. Check types, required fields, enums, patterns.
• Fix Broken JSON API — Fix broken JSON from LLMs: single quotes, trailing commas, missing braces, JS comments.
• JSON Schema Validate API — Validate data against JSON Schema (draft-07). Types, required, enum, min/max, pattern.
• Token Cost Estimate API — Estimate token count and USD cost for Claude, GPT-4o, Gemini. Essential for budget-aware agents.
• Webhook Send API — POST to any URL with any JSON payload. Agents need this to notify external systems.

View the full API catalog · Try in playground · Documentation