How do I automatically repair malformed JSON in an AI agent?

Steps

1. Install the Strale Python SDK, straleio.
2. Create the client with your API key.
3. Send the malformed string in the json field.
4. Read the repaired string from output["json"] or the parsed data from output["parsed"].

The solution

# pip install straleio
from straleio import Strale, DoRequest

strale = Strale(api_key="sk_live_YOUR_KEY")  # the SDK needs a key, even for free tools

broken_json = """{
  name: "John",
  age: 30,
  "hobbies": ["reading", "coding",],
  'active': true
}"""

result = strale.do(DoRequest(
    capability_slug="json-repair",
    inputs={"json": broken_json},
    max_price_cents=5,
))

print(result.output["json"])  # the repaired JSON string
print(result.output["parsed"])  # the same data, already parsed
print(result.output["fixes_applied"])

How it works

Language models often produce malformed JSON: trailing commas, single quotes, unquoted keys, comments and truncated output. The tool repairs these errors with fixed rules, so the same input always gives the same output.

The answer has json (the repaired string), parsed (the same data, parsed, so you do not need a second json.loads), valid and fixes_applied, a list that says what was changed, such as "Added missing closing brackets/braces". Severely malformed input may not be recoverable.

Tools in this guide