# Team Credence

- **Event:** [Built with Opus 4.7: a Claude Code hackathon](https://cerebralvalley.ai/e/built-with-4-7-hackathon)
- **When:** Apr 21 at 12:00 PM – Apr 27 at 2:00 AM (EDT)
- **Where:** Online
- **Team:** [LAKSHMI CHAKRADHAR VIJAYARAO](https://cerebralvalley.ai/u/Chakrivijayarao)
- **GitHub:** https://github.com/Lakshmi-Chakradhar-Vijayarao/credence-ai
- **Demo video:** https://www.youtube.com/watch?v=zKEf2k2bIsU
- **Gallery:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/103

Every day, developers trust Claude with decisions that matter. They share what they know — and what they don't. "The rate limit is probably around 50, I haven't confirmed it yet." A small admission. An honest one.

Then the session grows. Context gets compressed. And silently, without warning, RATE_LIMIT = 50 ships to production. The model just forgot you weren't sure.

We named this Epistemic Qualifier Loss (EQL), measured it across 50 compression scenarios, and built Credence to prevent it.

Credence is a five-layer epistemic enforcement system built natively into Claude Code. A sub-millisecond faithfulness probe blocks compression when uncertainty is present. A Truth Buffer re-injects unverified constraints before every generation. A Consistency Enforcer fires when a query overlaps a registered uncertain value. A Generation-Time Scanner annotates unverified literals before they ship. A Rust-native PreToolUse gate blocks irreversible tool calls — 98× faster than Python.

The Ghost Detector — powered by Opus 4.7 — catches implicit uncertainty with no hedging markers at all. Something no rule-based system can do.

As AI agents make higher-stakes decisions, uncertainty is not a weakness to hide — it is the most honest signal a system can carry. Credence is the first enforcement layer built around that principle.

22-tool MCP server. 178 tests. Built entirely with Claude Code.

---

Markdown version of https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/103. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt