User repository on GitHub
A minimal permission gate for Pi in the style of Claude Code's auto mode
- Forks
- 5
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Tools & CLIs
Pi Verdict
Permission gate extension for the pi coding agent: deterministic allow/deny rules and protected paths settle clear cases first, then gray-zone tool calls get one structured classify() verdict; Jev is an optional native classifier via TypeSafe direct, OpenRouter, or other gateways, failing closed on error.
- Stars
- 13
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Last verified
- Oct 2026
About this project
pi-verdict is a minimal permission gate for pi, inspired by Claude Code's auto mode: every tool call gets checked before it runs — allow, deny, or ask you first. Minimal — a 2k-line single-file core (the classifier rides pi's native classify() since 0.13) Built-in danger rules and your own allow/deny rules settle the clear cases first, at zero latency. Everything else goes to a model classifier that sees the conversation context. Any uncertainty or failure fails closed; nothing ever runs silently.
Quoted from the project's README. Read the full README.
- Languages
- TypeScript
- Platforms
- CLIMCPOpenRouter
- Capabilities
- Open sourceRuns locallyRequires API key
Highlights
- Set classifierModel to typesafe/jev-latest (or openrouter/typesafe/jev-1.13) to use Jev; requires pi >= 0.99, installed with pi install npm:pi-verdict.
- Three verdicts: ask routes ambiguity to a human and degrades to deny in non-interactive sessions; any uncertainty or classifier failure fails closed.
- Deny wins over allow, both beat the classifier, and the built-in deny floor is never overridden by the classifier or user allow rules.
- An optional confidence floor cascades low-confidence Jev verdicts to a fallback model; a demoted deny or ask can never be relaxed to allow.
- Codemode and MCP calls are gated too: a script may issue up to 256 nested calls, each adjudicated individually at roughly 350 ms per gray-zone call.
Watch out
MIT license; Jev support is an optional backend the README marks experimental, and tools exempted via ignoreTools lose the classifier's protected-path vigilance.
Questions about this entry
Short answers drawn from the linked project or page.
No. Built-in danger rules and your allow/deny rules settle clear cases at zero latency; only gray-zone calls reach a classifier, and Jev is one of several native classifier options in pi's catalog.
The gate fails closed: errors and timeouts deny, ambiguity becomes an ask in interactive sessions and a deny in non-interactive ones, so nothing runs silently without a verdict.
Install with `pi install npm:pi-verdict` (or `omp plugin install npm:pi-verdict`), then use `/automode on` or the `ctrl+shift+a` toggle. Auto mode is on by default, and the footer always shows `auto mode on` in green or `auto mode off` in yellow.
Set `classifierModel` to `typesafe/jev-latest` in ~/.pi/agent/config/pi-verdict.json. That opts into the native jev classifier, one structured classify() call per gray-zone verdict; otherwise the session model with thinking off is the default classifier.
No. Your allow/deny regex rules and the built-in danger rules settle the clear cases first at zero latency, and only what is left goes to the model classifier. `deny` wins over `allow`, and both beat the classifier.
No. The README describes it as a permission gate, not a sandbox, and says to stack OS isolation on top rather than treating the gate as a replacement. Its classifier allow is called an informed opinion, which is why a deterministic deny floor exists.
Version 0.13+ requires pi 0.99 or newer and runs on pi only. Older hosts, pi below 0.99 and oh-my-pi (omp), keep using the 0.12.x line from npm, whose classifier completion falls back to the pi-ai compat API on omp 18 while staying fail-closed.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Agent Handoff Gate: agents verify what they hand off
X: Jev as an LLM guardrail inside Agentgateway
X: Vercel's fx safety reviewer, measured at 18× faster
Reddit: Early experiments wiring Jev into Pi as a safety layer and model router.
More Jev resources like this
27GitHub stars
A DeepSeek Harness plugin where Jev Choice and Noul checks gate every non-read-only tool call - deny confident high-risk calls, escalate ambiguous ones, auto-approve clearly granted reversible ones - and optionally score session messages so valuable context survives instead of being dropped oldest-first.
Jev Tools & CLIs#community#plugin#guardrails
21GitHub stars
MIT-licensed Python tool router for coding agents that keeps obvious calls on a fast path, asks Jev (default typesafe/jev-1.13 via OpenRouter) for genuinely ambiguous choices, and can run bounded MCTS when multi-step consequences matter. Ships a CLI, MCP server and hook integrations.
Jev Tools & CLIs#community#routing#mcp
22GitHub stars
An MCP proxy and hooks adapter that screens an agent's tool calls before they run and the tool results before the agent reads them, with thresholds in a policy file.
Jev Agents & MCP#community#mcp#security
8GitHub stars
Supervisor for unattended coding agents: it tracks cross-turn state and asks one batched Jev Choice whether a trajectory should continue, pause or escalate, while deterministic code owns dangerous-command blocking, budget caps and Ed25519-signed receipts.
Jev Tools & CLIs#community#guardrails#safety
Community
VeridicalTech/EdwardPython
VeridicalTech/EdwardPython20.4kGitHub stars
Rust coding-agent harness (MIT, ~20k stars) whose crates/jcode-base/src/jev.rs is a typed TypeSafe Jev decision transport, used for memory pruning (memory_jev.rs), browser navigation (browser_jev.rs), and voice routing (voice_intent.rs) with separate provider selectors per purpose.
Jev Tools & CLIs#community#rust#coding-agent
Community
1jehuang/jcodeRust
1jehuang/jcodeRust3.4kGitHub stars
Vercel Labs' Unix-like coding agent CLI written in Zig, shipping a typesafe_permission_reviewer builtin that runs the agent's permission decisions through Jev rather than an LLM call. Model-agnostic and embeddable via libfx and ACP.
Jev Tools & CLIs#community#coding-agents#cli
Community
vercel-labs/fxZig
vercel-labs/fxZigFrom the Jev guides
Original write-ups on what this entry is for.
The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.
Jev Tools & CLIs#gateway#vercel#cloudflare
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.
Jev Tools & CLIs#classification#email#security
Read article
From the Jev community
Posts from builders shipping with Jev right now.