andreuvm/jev-reasoning-navigator
User repository on GitHub
Runtime supervision for autonomous AI agents
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
PRAXEON
Runtime supervision middleware for autonomous agents, published as andreuvm/jev-reasoning-navigator: no model-proposed action executes directly, and a deterministic runtime verifies preconditions, evaluates risk and submits actions to calibrated semantic judgment.
- Stars
- 1
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- andreuvm
- Last verified
- Oct 2026
Used inJev for Security & Privacy
- Languages
- Python
- Capabilities
- Open sourceRuns locally
Highlights
- No model-proposed action executes directly: the runtime issues an HMAC-SHA256 capability receipt with bounded validity before sandbox execution.
- Pipeline stages: context optimization, calibrated semantic judgment via TypeSafe or LAYA, operational policy, capability receipt and confined execution.
- Zero-config offline demo: clone, pip install -e '.[dev]', python demo.py, with no API cost, no registration and no external connection.
- Desktop app binds to 127.0.0.1 with a random per-run token and nine React views, from Live Canvas to Security and Settings.
- README badges report v1.1.0, Python 3.11+, 1117 passing tests, an MCP server and client, and providers TypeSafe, LAYA and CascadeRouter.
Quickstart
bash
git clone https://github.com/AndreuVM/praxeon.git
cd praxeon
pip install -e ".[dev]"
python demo.pyWatch out
Copy reviewed against the truncated Spanish-language repo README on 2026-10-15; the excerpt ends inside the interface description. GitHub metadata lists no license, and the TypeSafe, LAYA and CascadeRouter provider names appear only in badges in the visible text.
Questions about this entry
Short answers drawn from the linked project or page.
A supervision middleware and decoupled security runtime for autonomous LLM agents (ReAct, tool-use, tree-of-thought). The model proposes; the runtime decides what gets executed, intercepting every step before it reaches the environment.
When needed, each action is submitted to calibrated semantic evaluation through providers such as TypeSafe or LAYA, separate from operational policy, before a signed capability is issued and the tool runs confined in a sandbox.
Yes. The quick start is fully offline and zero-config: clone the repo, pip install -e .[dev], then python demo.py for a visual demo, or python scripts/run_benchmarks.py for the reproducible benchmark suite. No registration or external connection is needed.
A signed HMAC-SHA256 capability with bounded validity is issued for the action; the runtime then executes the tool confined inside an isolated sandbox, so a proposed action never has direct authority over the environment.
praxeon app starts a local server only on 127.0.0.1, protected by a random per-run token known only to the window, and opens a native window via pywebview, or --browser without it. Data survives restarts in .praxeon_cache.
Nine React views: Live Canvas, Sessions, Decisions, Agents, Workflows, Providers, Tools (MCP), Security and Settings. The README excerpt covers the workspace selector before truncating, so the remaining views are known by name only.
Python 3.11 or newer; the README badges report version v1.1.0 with 1117 tests passing, an MCP server and client, and TypeSafe, LAYA and CascadeRouter providers. The repository is named andreuvm/jev-reasoning-navigator.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
YouTube: Kubernetes admission control with a decision model, fail-closed.
Hacker News: SafeDep Gryph plus Jev checks for compromised coding-agent behavior.
2 pointsDetecting Compromised AI Coding Agents with Jev and GryphRead the thread on Hacker NewsX: Jev and Laya on security queries
X: Testing Jev for prompt injection
More Jev resources like this
Lets AI agents use Windows applications through MCP or the command line: hand off a bounded task and Jev observes windows, chooses controls, types supplied text and navigates locally, returning a result or a request for help.
Jev Agents & MCP#community#catalog-import#logicrw
Community
jacks3tr/jev-desktopPython
jacks3tr/jev-desktopPythonA local Python service that keeps your skill catalog out of the main model's context: it reads your SKILL.md library, asks Jev which skills fit, and returns only the selected instructions through a single read-only MCP tool, with Claude Code, Codex, and Cursor adapters.
Jev Agents & MCP#community#catalog-import#logicrw
Community
himomohi/jev-skill-routerPython
himomohi/jev-skill-routerPythonA Hermes Agent skill that turns an intention into a written finish line: machine-checkable requirements run as shell commands, fuzzy ones go to Jev for calibrated probabilities, and exit 1 names the weakest unmet requirement.
Jev Agents & MCP#community#catalog-import#logicrw
Official MCP server for JevGuard and TypeSafe AI: zero-dependency guardrails over JSON-RPC 2.0 on stdio, gating shell commands and code patches, caching decisions in local SQLite and pruning state.
Jev Agents & MCP#community#catalog-import#cobanov
Community
seb4ez/jevguard-mcpPython
seb4ez/jevguard-mcpPythonAn open-source Jev-based MCP router: it discovers, catalogs, classifies and deduplicates every MCP server and tool you have, then exposes each agent only the small, authorized subset it needs for the task. Routing decisions are System One Choice/Score/Noul questions, scored by a pluggable engine at the edge or hosted.
Jev Agents & MCP#community#catalog-import#logicrw
Community
amantux/mcp-routerPython
amantux/mcp-routerPythonAn 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
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.