Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Jev Policy Engine

Python policy-as-code SDK that runs Noul, Choice and Score rules defined in YAML through Jev, with audit mode and fail-closed options.

Libraryprototype
View on GitHub ↗

Overview

Jev Policy Engine is a policy-as-code SDK for Jev. Teams define governance rules as Noul, Choice and Score questions in YAML, tag them (for example cpq or security), and run only relevant rules against an input. Rules support an audit (shadow) enforcement mode and an on_failure: fail_closed setting. Rules are loaded through a YAML provider designed to be extended to sources like Notion, CSV or Salesforce custom metadata.

The README describes four uses: gating agent tool calls such as run_command or execute_sql and returning a block so the agent self-corrects; routing tasks by a complexity score to expensive or cheap agents; enforcing rules in CI by piping a diff or plan into the CLI; and validating form payloads in backend middleware with a 403 on violation. It works with the TypeSafe SDK and with Vercel AI Gateway.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes this proposed command or payload violate the policy?(inferred)yes · no
ScoreHow complex is this request (1-10)?—
ChoiceWhich category does this input fall under for the policy?(inferred)—

How it works

  1. Write rules in YAML with tags and enforcement mode
  2. Pass an input (tool args, diff, form payload) to the engine
  3. Jev answers the tagged rules
  4. Return pass, block or audit-only results; fail closed on errors if configured

Strengths

  • Non-developers can manage rules in YAML
  • Audit mode for shadow testing
  • Usable in agents, CI and backends

Caveats

  • Semantic rules are probabilistic despite 'deterministic' framing
  • Requires the TypeSafe SDK and key
  • Early project

Getting started

git clone https://github.com/bhavinm/jev-policy-engine.git && cd jev-policy-engine && pip install -e . && pip install typesafe-sdk pyyaml

At a glance

Type
Library
Status
prototype
Language
Python
Runtime
Library and CLI
Has code
yes

Use cases and integrations

How curated lists describe it

  • Policy conformance: universal Policy-as-Code SDK that allows DevOps and security teams to enforce deterministic AI governance rules in YAML via Jev with audit mode and fail-closed controls.

    yibie/awesome-jev

Frequently asked questions

How are rules defined?

As Noul, Choice and Score questions in YAML, with tags and enforcement settings.

Can I test rules without blocking?

Yes. enforcement_mode: audit runs rules in shadow mode.

Can it run in CI?

Yes. Pipe a diff or plan into jev-policy evaluate with a rules file.

Related projects

  • jev-zig-cli

    Auditable Zig terminal agent pairing Codex for coding with Jev typed preflight and postflight checks under deterministic policy gates.

    CLIproductionZig
  • actiongate-jev

    Gateway that checks agent tool calls with policy plus Jev evidence, then binds approval to a single-use permit for that exact action.

    MCP serverbetaTypeScript
  • Agent Chaperone

    MCP proxy and hooks that screen agent tool calls before they run and their results before the agent reads them, using Jev judgments.

    MCP serverbetaTypeScript
  • agent-guardrail-jev

    Xano backend template where Jev judges each agent tool call as safe, needs review or dangerous, and uncertain calls wait for a human.

    ProjectbetaXanoScript
  • approval-judge-bridge

    OpenAI-compatible endpoint that answers an agent's command-approval call with APPROVE, DENY or ESCALATE from a Jev Choice, failing closed.

    ProxybetaPython
  • claude-jev-plugin

    Claude Code plugin adding Jev judgments to Bash, Write and Edit calls and their outputs; a semantic guardrail, not a sandbox.

    Pluginbeta