Inspekt is a TypeScript-native proxy that diagnoses API errors in real-time. Instead of just returning raw data, it uses AI to analyze headers, status codes, and bodies to provide actionable fixes for 4xx/5xx errors
I’m Jamaldeen, and I built Inspekt because I got tired of "log-diving" every time a 4xx error popped up. Most debuggers just show you the raw data; I wanted something that told me the solution.
Inspekt isn't just a status-code checker. It’s a semantic proxy. By intercepting the full request/response cycle (headers + body), it uses AI to diagnose exactly why a request failed even in tricky scenarios like GraphQL returning a 200 OK with an error payload inside.
Key features to test:
Deep Packet Diagnosis: It analyzes the full JSON body and headers to find the "invisible" bugs.
Security Audit: It automatically flags missing CORS headers, insecure routes, or accidental credential leaks.
Context-Aware Fixes: It doesn't just say "403 Forbidden"; it tells you which specific permission or header is missing from your request.
Smart Truncation: It handles massive payloads by intelligently trimming data to keep your AI tokens low while keeping the core logic intact.
Roadmap based on today's feedback:
Local Privacy Scrubbing: Stripping PII/Auth tokens locally before they ever hit the AI.
Custom Error Triggers: Allowing you to define what counts as a "failure" even on 200-status responses
Thank you for all the incredible technical questions so far. I’ve been reading every single comment, and based on your feedback regarding PII and edge cases, I just pushed Update 01 to the live proxy.
Here is a quick FAQ to cover the most common questions and the new security features
1. How do you handle sensitive data (API Keys/Tokens)?
We just pushed a live update! Inspekt now uses Local Privacy Scrubbing. Before your request leaves our server, we automatically redact headers like Authorization, Cookie, and X-Api-Key. Your secrets stay on the proxy; only the metadata goes to the AI.
2. Does it support GraphQL?
Yes! Because Inspekt analyzes the semantic meaning of the response body (not just the HTTP status code), it correctly identifies errors even when a GraphQL server returns a 200 OK with an error array inside.
3. What happens with massive JSON payloads?
We use Smart Truncation. We intelligently trim large bodies (8,000+ chars) to keep AI token costs low while preserving enough structure for the model to diagnose the issue accurately.
4. Can I self-host this?
Absolutely. The repo is open-source. You can clone it, plug in your own OpenRouter and Upstash keys, and run it on your own infrastructure for total control.
5. What’s next on the roadmap?
Custom Success Boundaries: Define your own rules (Latency, Schema requirements) for when the AI should trigger.
NPM Middleware: A one-line integration for Express and Next.js apps.
Jamaldeen
Hey Product Hunt! 👋
I’m Jamaldeen, and I built Inspekt because I got tired of "log-diving" every time a 4xx error popped up. Most debuggers just show you the raw data; I wanted something that told me the solution.
Inspekt isn't just a status-code checker. It’s a semantic proxy. By intercepting the full request/response cycle (headers + body), it uses AI to diagnose exactly why a request failed even in tricky scenarios like GraphQL returning a 200 OK with an error payload inside.
Key features to test:
Deep Packet Diagnosis: It analyzes the full JSON body and headers to find the "invisible" bugs.
Security Audit: It automatically flags missing CORS headers, insecure routes, or accidental credential leaks.
Context-Aware Fixes: It doesn't just say "403 Forbidden"; it tells you which specific permission or header is missing from your request.
Smart Truncation: It handles massive payloads by intelligently trimming data to keep your AI tokens low while keeping the core logic intact.
Roadmap based on today's feedback:
Local Privacy Scrubbing: Stripping PII/Auth tokens locally before they ever hit the AI.
Custom Error Triggers: Allowing you to define what counts as a "failure" even on 200-status responses