GH·RADAR

Mask secrets in logs before an LLM reads them and put them back in the answer, streamed replies included. Keyed HMAC placeholders, Python stdlib, one file.

The Nmap for cloud storage. Find exposed S3, GCS and Azure buckets, fast.

An MCP server for Wallos, the self-hosted subscription tracker. Runs locally over stdio, reaches instances on your LAN, and never sends your API key anywhere else.

Local-first LLM routing with quality gates, automatic fall-through, and a JSONL usage ledger for cost attribution - Python 3.11+ stdlib only.

Turn any public GitHub repository into a single LLM-ready markdown digest - smart filtering, priority ordering, and a token budget that never truncates silently.

Scans a Node/TypeScript repo for personal data, maps each finding to LGPD and GDPR articles and drafts RoPA/DPIA documents. Deterministic engine; the LLM only labels ambiguous fields and drafts prose. Not legal advice.

Is your probability forecast honest? Brier, ECE and reliability, one temperature fitted on held-out days with day-clustered bootstrap CIs, and an error-budget cut on the gate's own scale. Stdlib Python.

Runnable examples that trace, evaluate, and debug LLM apps with Opik — 22 small apps across Python, TypeScript, Java & n8n (OpenAI, Anthropic, LangChain, LangGraph, LlamaIndex, DSPy, Temporal, Vercel AI SDK, Mastra, and more).

A Claude Code mod that docks a live pane beside your conversation: what Claude is doing right now, every agent, workflow and background task as a process you can open, how fast it is writing, and every prompt you sent, one click from its place in the transcript

Inspect HAR files locally, triage network issues, and export a reviewed sharing pack with sanitized HAR, issue draft, and audit.

Message broker for AI coding agents (Claude Code, Codex, OpenHands) across machines: atomic claim, loop guards, read receipts and sender alerts.

Local compute guard for a shared PC: thread/load budget, night queue, stepped thermal guard (EN/pt-BR)

AI agent that researches any company on the web and returns a one-page sales brief with cited sources in ~1 min. Next.js + OpenAI Responses API + SQLite. Built document-first with AI coding agents: specs → plan → one commit per step.

A free, open-source Mac cleaner that shows exactly what it removes. Native SwiftUI, no network, Trash by default.

LLM-assisted accessibility fixer with a deterministic verification loop: the model only proposes values, code builds bounded patches and a real-browser re-audit decides. Flags every LLM value for human review.

See your codebase. Fit your AI context. A browser-local visual context builder with token budgets, secret redaction and zero dependencies.

Open Agent Mission & Control Platform — connect your own AI (Claude Code), watch a small agent team work, approve or block every action before it runs. Local-first, no account, no telemetry. Apache-2.0.

Claude Code skills by Eugenio Nerelli: pingpong, a build-critique-repair-test loop with subagents

Compact Claude Code status line: model, 5h/7d rate-limit bars, context-window bar, plugin badges

Dependency-free MCP server for creating, validating, and previewing Айрен .itx tests

Append-only event bus for AI agents — hash-chained tool/decision log, one-command replay + side-effect audit.

Translate comic pages: OCR of the balloons, AI-assisted translation and review, text redrawn on the page

AI agent for hotel guest emails: finds every request, checks booking, availability and policy, drafts a cited reply for staff approval. Simulated engagement, synthetic data.

ESP32-S3 智能鱼缸监测终端：LVGL 触屏界面 · 双工语音对话 · 摄像头 · TF 卡相册 · MQTT 上云 · 云端 AI 工具调用（ESP-IDF v5.5.3）