CodeJury

CodeJury — Independent AI Code Verification AI writes code faster than teams can review. CodeJury is a jury of Fetch.ai agents that independently verify every pull request (PR) before it merges. The key difference: Tests are generated from the specification BEFORE the code is read. This ensures the verifier cannot share the implementer's blind spots. HOW IT WORKS Paste a GitHub PR URL into ASI:One. The orchestrator reads the diff, fetches the specification, and routes the review to specialized agents. 1. Correctness Agent: Always runs. Generates tests for edge cases, logic, and contracts using only the specification. 2. Security Agent: Activates for authentication and token-related code. Checks for IDOR, injection, and authentication bypass across 7 attack surfaces. 3. Verdict Agent: Applies a deterministic policy gate. Critical defects result in BLOCK. The LLM writes the summary but never decides the verdict. DEMO RESULTS 27 tests generated 4 defects detected 2 critical IDOR vulnerabilities — Any user could access another user's private data. 2 major performance bugs Both critical vulnerabilities were independently discovered by the Correctness and Security agents, confirming the findings from different angles. Final Verdict: BLOCK — 90/100 INDEPENDENCE BY DESIGN Spec-first testing: Tests are generated before implementation code is examined. Separate agent contexts: Each agent evaluates independently. Deterministic verdict: Policy-based decisions are protected against prompt injection. FETCH.AI TECH STACK 4 uAgents deployed on Agentverse Agent Chat Protocol for agent communication ASI Interactive Cards with custom summary and carousel report ASI:One for agent discovery and interaction Complete verification workflow in one conversation. No frontend required. Built by Manali Gawande & Siddharth Jetling AIE Code Summit 2026