jev-tools
Six evidence-oriented Jev tools for pi, omp and any MCP client: ask about code, find files, locate ranges, check diffs and select tests.
Overview
jev-agent-tools provides six tools for coding agents that use the Jev API format to answer typed questions about repository evidence. jev_ask combines a note, files, earlier versions or command output into one judgment; jev_ask_files applies the same questions independently to each candidate file; jev_find_files finds an entry point for a behavioural goal when the filename is unknown; jev_locate_in_file finds the relevant range in a large file; jev_check_diff reviews completed changes for risk, documentation or specification issues; and jev_select_tests picks commands for affected existing tests without running them.
The tools are meant to complement reading, searching and execution, not replace them. They install as a pi package or omp plugin, and since 0.2.0 a stdio MCP server exposes the same tools to Claude Code, Cursor, Codex CLI, VS Code and others; releases are in the official MCP Registry. Commands inside jev_ask run with ordinary shell permissions and can be disabled. The project includes a getting-started guide, per-tool docs and agent instructions for CLAUDE.md or AGENTS.md.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this file contain the entry point for the stated behaviour?(inferred) | — |
| Score | How risky is this completed change?(inferred) | — |
| Yes / no | Is this existing test affected by the change?(inferred) | — |
How it works
- Agent calls a jev_* tool with a question and evidence root
- Tool gathers files, diff or command output as state
- Jev returns typed answers
- Agent reads or runs what the tool points to
Strengths
- Works across pi, omp and any MCP client
- Narrow tools for navigation, review and test selection
- Listed in the official MCP Registry
Caveats
- jev_ask commands run without an extra sandbox unless disabled
- Requires a Jev-compatible endpoint and key
Getting started
pi install npm:jev-agent-tools@0.4.0, or add `npx -y -p jev-agent-tools jev-agent-tools-mcp --root /path/to/repo` as an MCP server.At a glance
- Type
- Service
- Status
- beta
- Language
- TypeScript
- Runtime
- Pi/omp extension and MCP server
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Six evidence-oriented tools for pi and omp coding agents, compatible with the Jev API format.
logicrw/awesome-jev-projectsSix evidence-oriented tools for pi and omp coding agents, compatible with the Jev API format.
hellogumbo/awesome-jev
Frequently asked questions
What are the six tools?
jev_ask, jev_ask_files, jev_find_files, jev_locate_in_file, jev_check_diff and jev_select_tests.
Do they replace reading code?
No. The README says they complement reading, searching and execution.
Can I use them outside pi?
Yes, via the stdio MCP server with any MCP client.
Related projects
Gives Claude Code, Codex, Pi and OpenCode five typed Jev tools (classify, check, score, rank, ask) plus a skill on when to use them.
MCP serverbetaTypeScriptSemantic grep for code, diffs and CSV rows that packs 16 chunks per Jev request and prints file:line hits with CI-friendly exit codes.
Agent skillbetaTypeScriptCLI that scores every test against a git diff with Jev and emits filter arguments for vitest, jest, node, Playwright, cargo, go and rspec.
ProjectbetaTypeScriptCLI, MCP server and Claude Code plugin for closed-form yes/no, choice, score and rank decisions via Jev, OpenAI Decisions or Claude.
MCP serverbetaTypeScriptGo CLI that gives coding agents Jev-powered lookups and calibrated yes/no or choice verdicts instead of reading whole files.
CLIbetaGoMCP server that finds relevant code with Jev Choice and Noul checks instead of grep sweeps, and audits diffs against task intent.
MCP serverprototypePython