What problem does JEV Fuse solve?

Raw calibrated probabilities in front of production agents invite razor-thin boundary hazards: with a naive P above 0.50 cutoff, 0.51 and 0.49 can produce catastrophic unintended actions. JEV Fuse turns those numbers into deterministic, fail-closed policy actions.

What actions can it produce?

ALLOW, ASK, DENY, KEEP, TRUNCATE or DROP. A reverse proxy sits between callers like Claude Code, Cursor/MCP and the official Python SDK, and decision engines, translating each response into one governed action.

How does it handle shell commands?

A 4-tier AST guard gate parses commands instead of pattern-matching them, catching constructs like find . -exec, subshells, backticks and redirects that regex-based checks miss.

What is stored in the audit trail?

A write-ahead log on DuckDB keeps a persistent decision log so you can replay why an autonomous system allowed, blocked or altered an execution, which the README frames as the verifiable audit trail the layer adds.

How does it protect against rate limits?

A singleflight component deduplicates duplicate concurrent evaluations, so bursty agent loops do not exhaust upstream quotas or pay for repeated identical calls with unnecessary latency.

What does it speak on the wire?

The SystemOne contract at POST /v1/systemone, so it is a drop-in proxy for TypeSafe Jev and local Laya engines. It ships on PyPI as jev-fuse v0.1.0 and needs Python 3.12 or newer.

Who built it and what are the limits?

0xshikhar's project under Apache-2.0, described as an independent governance runtime. The README excerpt covers the problem statement and architecture but no quickstart, and the under-35 ms latency figure refers to the decision models rather than the proxy path.

Jev directory

CommunityJev Integrations & Adapters

JEV Fuse

A governed execution layer and drop-in proxy between typed-decision models like Jev and callers such as Claude Code, Cursor/MCP and SDKs: raw probabilities become deterministic policy actions from ALLOW to DENY with a DuckDB audit trail.

See highlights Open github.com

Stars: 2
Language: Python
License: APACHE-2.0
Reading time: 2 min
Published by: Community
Author: 0xshikhar
Last verified: Oct 2026

0xshikhar/jev-fuse

User repository on GitHub

View on GitHub

A governed execution layer between typed-decision models (TypeSafe Jev) and agent/AI tooling like Claude Code, MCP, and AI SDKs, turning probabilistic decisions into deterministic, policy-controlled, and auditable actions.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw governance proxy audit

Works forJev for Decision gating Jev for Realtime control Jev for Verification

Used inJev for Security & Privacy Jev for Software & DevTools

Languages: Python
Platforms: MCP API DuckDB
Capabilities: Open sourceSelf-hostedRequires API key

Highlights

Translates a raw model probability into one of six actions: ALLOW, ASK, DENY, KEEP, TRUNCATE or DROP.
A 4-tier AST guard gate catches shell parser bypasses such as find -exec, subshells, backticks and redirects.
Singleflight deduplication collapses duplicate concurrent evaluations, protecting upstream rate limits in fast agent loops.
A WAL audit on DuckDB keeps a persistent decision log that can explain why a system allowed, blocked or altered an execution.
Speaks the SystemOne wire contract at POST /v1/systemone and ships on PyPI as jev-fuse v0.1.0 for Python 3.12+.

Watch out

Grounded in the truncated repository README and GitHub metadata reviewed on 2026-10-15; the README was cut off mid-diagram below the architecture, so quickstart and configuration sections were not reviewed. Apache-2.0 licensed, and the under-35 ms figure refers to decision models, not the proxy itself.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

Decision gates in agent frameworks: what each integration ships

Oct 16, 2026

How Pydantic AI, Mastra, TanStack AI, DSPy, Spring AI, and LiteLLM expose Jev: where the typed decision sits in the loop, what each integration returns, and which big frameworks are only adjacent.

Jev Integrations & Adapters#agents#frameworks#integrations

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai