jev-tools
An open-source toolkit for verifying AI coding agents. It provides six repository forensics tools that collect evidence from code repositories and submit it to an independent judgment mechanism to help agents verify their conclusions, with support for MCP protocol integration.
Timeline (last 90 days)
NomenAK has open-sourced jev-tools, a set of tools for letting AI coding agents gather repository evidence to verify their work.
The core design of jev-tools is to separate "evidence gathering" (collecting evidence from the repository) from "judgment" (answering specific questions based on evidence) and assign them to different mechanisms.
jev-tools currently provides six tools: jev_ask, jev_ask_files, jev_find_files, jev_locate_in_file, jev_check_diff, and jev_select_tests.
jev_ask answers focused questions by assembling evidence from multiple sources.
jev_ask_files evaluates the same question independently across multiple files to avoid misjudgments from a single-file perspective.
jev_find_files locates relevant files based on a behavior or development goal.
jev_locate_in_file pinpoints relevant information within a source file.
jev_check_diff checks code changes for potential inconsistencies and issues.
jev_select_tests identifies existing tests related to modified code.
Once integrated, jev-tools can detect whether documentation has become outdated after code changes.
1 more timeline events
All Facts (11)
jev-tools cannot eliminate hallucinations, and a positive judgment does not constitute proof of correctness, because the evidence may be incomplete or the judgment itself may be wrong.
50%UnverifiedThe core design of jev-tools is to separate "evidence gathering" (collecting evidence from the repository) from "judgment" (answering specific questions based on evidence) and assign them to different mechanisms.
50%Unverifiedjev-tools currently provides six tools: jev_ask, jev_ask_files, jev_find_files, jev_locate_in_file, jev_check_diff, and jev_select_tests.
50%Unverifiedjev_ask answers focused questions by assembling evidence from multiple sources.
50%Unverifiedjev_ask_files evaluates the same question independently across multiple files to avoid misjudgments from a single-file perspective.
50%Unverifiedjev_find_files locates relevant files based on a behavior or development goal.
50%Unverifiedjev_locate_in_file pinpoints relevant information within a source file.
50%Unverifiedjev_check_diff checks code changes for potential inconsistencies and issues.
50%Unverifiedjev_select_tests identifies existing tests related to modified code.
50%UnverifiedOnce integrated, jev-tools can detect whether documentation has become outdated after code changes.
50%UnverifiedNomenAK has open-sourced jev-tools, a set of tools for letting AI coding agents gather repository evidence to verify their work.
50%