User repository on GitHub
Local Jev / System One–compatible gateway — not TypeSafe’s Jev. Fast, privacy-first, offline.
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Reflex Gate
Local Windows gateway that fronts agent loops with stop, extract, and scan endpoints plus a System One-shaped typed decision path. Runs a small coder model over llama.cpp at localhost with telemetry and self-tests.
- Stars
- 6
- Language
- Go
- License
- MIT
- Published by
- Community
- Author
- jagsan-cyber
- Last verified
- Oct 2026
About this project
Ultra-Fast, Resilient Local AI Gateway for Autonomous Agent Loops. ReflexGate is an open-source project created to drastically accelerate and stabilize local AI agent execution loops with zero token costs and minimal latency. If ReflexGate helped speed up your local agent workflows, saved your API token expenses, or streamlined your LLM pipeline, please consider buying a coffee to support continued development and maintenance!
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- Go
Highlights
- Three loop endpoints for stop gating, JSON extraction, and safety scanning over log text.
- Compatible typed noul and choice path for local triage before cloud calls.
- Windows GUI with telemetry, tray residency, and kernel-bound server cleanup.
- Self-test suite with trap patterns plus offline reuse after first download.
Watch out
Windows-only desktop build with local-model accuracy limits; hard cases can still lose to the hosted API. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Use it as local-first triage before any cloud worker. The typed path accepts model plus state plus questions with instructions and returns noul or choice answers, but it is explicitly not a distillation and not affiliated with TypeSafe.
They take a single log string for loop-exit gating, status and error extraction, and secret plus crash scanning. Typed noul and choice decisions use the separate POST /v1/systemone path and must not receive only a log payload.
Download the executable from Releases, run it, use auto-fetch for the model and matched llama-server binary, then start the server for a localhost gateway. After that download everything runs offline with no outbound requests and assets kept locally.
They are trap-pattern checks, not general guarantees. Quick mode runs a smaller case set and thorough mode a larger one across stop, extract, and scan categories, and you should rerun the suite on your own GPU or CPU because latency scales with hardware.
Windows 10 or 11 x64, about 2 GB of VRAM at the 8,192-token max context on a GPU backend or about 4 GB of RAM on CPU, and roughly 2 GB of disk for the model and llama-server binary, which are downloaded on first run and kept for offline use.
The bundled llama-server.exe is bound to a Windows Kernel Job Object with JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE, so closing or killing ReflexGate terminates the model process automatically instead of leaving an orphan running.
Quick runs a smaller case set and Thorough a larger one across the stop, extract and scan categories, over 13 production trap patterns such as fake success and secret leaks. The README says to rerun the suite on your own GPU or CPU because latency scales with hardware.
More Jev resources like this
16GitHub stars
Shell wrapper that intercepts unknown subcommands, reads the target CLI help output, and asks Jev which documented subcommand matches the intent behind the typo.
Jev Agents & MCP#community#jevbest#catalog-import
Community
syumai/jevyoumeanGo
syumai/jevyoumeanGo40GitHub stars
Claude Code plugin with jev find, jev ask, and a Read hook that keeps large-file context out of the agent by scoring relevance with Jev.
Jev Agents & MCP#community#jevbest#catalog-import
Community
BorisLeMeec/jevGo
BorisLeMeec/jevGo3GitHub stars
Small Go triage tool that sends each message plus three questions to Jev in one request, then prints an urgency-sorted table, a low-confidence review list, and an HTML report.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
boldbug1/jev-triageGo
boldbug1/jev-triageGo3GitHub stars
Go MCP server exposing Jev through OpenRouter or TypeSafe as five tools: decide, classify, verify, rerank and list_models. Each tool call is one paid inference request, answers are validated against the question asked, and the unchanged provider response is returned with its usage cost.
Jev Agents & MCP#community#jevbest#catalog-import
Community
freepik-company/jev-mcpGo
freepik-company/jev-mcpGo1GitHub stars
Archived Go proxy that sat between a coding-agent CLI and the upstream API to rewrite tool schemas and history with Jev, kept for the negative result: tool steering cut tokens on one model and added 37% and 51% on two others.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
nekowasabi/jev-routingGo
nekowasabi/jev-routingGo1GitHub stars
A Go harness for System One decision models, covering hosted Jev through the Vercel AI Gateway or the native API plus Kev running in-process on llama.cpp, with typed helpers for gating decisions and a CLI for trying a state and questions from a shell.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
taigrr/gojevGo
taigrr/gojevGoFrom the Jev community
Posts from builders shipping with Jev right now.