Hey everyone. Built Shim after 6 months debugging production AI agents. Same failure pattern: LLM returns invalid JSON, app crashes. **The problem:** - GPT-4 stream cutoffs (incomplete objects) - Claude markdown wrappers (```json fences) - Unquoted keys, trailing commas, NaN values - Broken production apps **The usual fixes don't work:** - Rewrite prompts 47 times → still breaks - Add 200 lines of error handling → still breaks - Parse with regex → please don't **How Shim works:** 1. Point LLM output at api.shim.so 2. We intercept and repair malformed JSON 3. Return valid JSON to your app 4. Sub-50ms latency **Zero-persistence promise:** No raw content stored. Ever. Only metadata (repair count, confidence). Check our DB schema: github.com/shimhq/shim-core/blob/main/db/schema.sql **Streaming support:** Push chunks as they arrive. Get valid JSON without waiting for EOS. **Pricing:** 1,000 free repairs/month. No credit card. $19/month after. Try it: shim.so/demo npm install shim-sdk pip install shim-sdk What JSON nightmares are you dealing with?

Shim.so

JSON repair API for broken LLM outputs. Sub-50ms streaming.

JSON repair API for broken LLM outputs. Sub-50ms streaming.