Code agents

The official open-source repository for community-built prototypes and tools powered by the Oxlo API

LAMA CARAVAN — Local And Multi-backend Adaptive Computation And Routing Across Various AI Nodes. Self-hosted control plane for llama.cpp fleets and AI agents: hybrid local/cloud token routing with queues & schedules, remote server cells, HF model browser, usage accounting. Stdlib Python.

Credits-first router for agent fleets, plus a lifeguard for long Claude Code sessions. If it runs in bash, it can be orchestrated. OpenClaw, Hermes, Claude Code, Codex, Grok.

The reachpad CLI: workspaces where Claude, Codex and other coding agents build full-stack apps and share them by link, from your terminal.

Framework-agnostic Roblox/Luau best-practices skill for AI agents. Performance, security, data, platform limits, Studio MCP safety, and 22 system blueprints. Installs to 62 agents including Claude Code, Cursor, and Windsurf.

Your backlog, shipped in parallel — a desktop app that runs Claude agents across your repo's tickets, each in its own git worktree. Triage in, PRs out.

Review-first UI/UX design intelligence for coding agents: plan, route, implement, and verify intentional product experiences.

Most tools check the code your agent wrote. nullius checks how it got there — citations, session journals, weakened tests.

Parallel model rotation across any CLI. Auto-discovers models, health-checks in <2s, scores by capability+cost, picks the best for every task.

Governed software development for AI coding agents.

90,896 Hermes Agent skills — 79k+ shipped as COMPLETE offline folders (SKILL.md + scripts, references & assets). Instant offline search across ClawHub, skills.sh, LobeHub, gstack & more.

A fast multi-threaded Instagram username availability checker with Telegram notifications. Hunt for rare usernames with random generation or batch checking from lists.

AI-powered marketing automation toolkit built on Claude Code subagent orchestration. Autonomous workflows for content creation, campaign management, SEO, lead generation, and browser automation.

Turn goals, existing Skills, and authorized work records into bounded Agent Loops, then export a Codex Skill or Compact Prompt with separate verifiable evidence.

Provider-neutral email review and future-mail cleanup for AI agents

Sleep-time memory consolidation for Claude Code agents, with a cross-project shared-consciousness layer.

Agent skill que gera e mantém o panorama-atual.md, o arquivo-base de status do projeto (fase atual, implementado, pendente e decisões). Compatível com Claude Code, OpenCode, Cursor e Codex.

Web-first Hermes skills index and safety gate: find SKILL.md packs before you install them.

Bridge Linear agent sessions to a local agent runtime

Run a live Meta (Facebook/Instagram) ad account end to end with Claude Code. 6 skills + MCP setup. You approve every spend. Free and open source.

Steel (steel.dev) cloud browser provider plugin for Hermes Agent

Testing and evaluation framework for AI agents

BonoBox (BBX): useful tools for AI agents and everyday automation — real tests, clear limits.

A swiss knife for your Agents to browse with no limits

Claude Code Chrome extension — runs the real Claude Code CLI locally, with the chat in your browser's side panel. Claude gets browser tools over MCP (DOM, console, network, screenshot, click, type) plus an on-page design toolkit. Local only, no telemetry.

Geocode places, fetch global weather forecasts, ERA5 historical climate, marine conditions, air quality, and terrain elevation via MCP. STDIO or Streamable HTTP.

Read-only-first Codex skill for auditing and refactoring existing skills

Claude Code integration plugin for DeepSeek Harness, providing native Claude conversations powered by the Claude Agent SDK, with approvals, session continuity, and DSH tool support.

A local MCP server for safely controlling the ChatGPT web UI through a persistent browser profile.

Self-hosted email CLI / API / MCP server for AI agents and apps

高置信度命理 Skill —— 八字/紫微斗数/七政四余三盘互译的命理对话伴侣（Claude Code / Codex / Kimi 等支持 SKILL.md 的 agent CLI 通用）：置信度三道题 + 斯多葛语气 + 收束自校准，支持缺盘降级，数据全本地。A Chinese metaphysics companion for agent CLIs with three-chart cross-translation (BaZi/Ziwei/Qizheng), confidence-gated interpretation, stoic voice, and self-calibrating local memory.

Multi-agent LLM peer review: an editorial board for manuscripts. Eight specialist reviewers, an advocate/skeptic debate, an author rebuttal, and an editor's verdict, with full reports at every stage.

DSH 协作模式 preset：双停点确认、版本化状态、候选冻结与独立审查的多 Agent 协作规范 · Governed multi-agent collaboration preset for DSH

Connect Claude Code, Cursor, and Codex to Workflow (workflow.games): your AI coding agent plans requirements, files bugs with dedup, runs real online QA, and drives ticket status — every write verified by read-back. Agent Plugins 1.0.0 + Claude Code marketplace.

Grounded perceptual memory for personalized agents: VLM grounding, specialist identity encoders, and a training-free in-model attention bank. Paper: arXiv:2608.28609

LLM API-Powered Job Listings Data Cleaning Tool: Showcase for Data Scientists

Monitor Codex usage and switch accounts across Codex CLI and desktop apps on macOS and Windows.

Live token usage & cost HUD inside Codex App. Renderer-injected, fully local, zero telemetry.

Live Codex quota tray panel with animated weather scenes for Windows

中国大陆普通人三分钟用上 Claude Code——BYOK、无账户、不经手你的数据。MIT 开源。

Isolated Claude Code profiles via CLAUDE_CONFIG_DIR. No credential vault swap.

Windows system-tray monitor for Claude usage limits - live session/weekly meters, usage graph, alerts. Unofficial.

类脑联想记忆库：SQLite + bge-m3 + 扩散激活的单机记忆引擎。纯向量回答'哪条最像'，联想记忆回答'这整件事是什么'。Brain-inspired associative memory engine (spreading activation, online learning, flexible forgetting).

Self-hosted, multi-user transcription platform: record or upload audio. Speaker-labeled, timestamped transcripts, Recognize speakers across recordings, Summarize, extract action items and chat over your transcripts with your own OpenAI-compatible LLM. Your Audio, your Server, your Model. Tested: Laptop RTX4070, RTX3090, RTX5090 and ROCM Strix Halo

Two Claude Code skills: /showoff turns any project, PR, or app into a cinematic evidence-grounded showcase video -- and /figureout gives Claude the ability to watch, transcribe, and understand any video or screen recording.

Production-shaped control plane for AI support agents: scoped routing, policy broker, tool boundaries, approval queue, replay verification, audit traces, and human handoff. Prototype on synthetic/sample data — no production users.

An AI SOC analyst as an agentic skill — investigates and triages Exabeam New-Scale alerts end to end, with governance gates and guardrails

Deterministic observation-integrity contracts for AI-agent tool results. HTTP 200 is not an observation.

AI-powered WhatsApp receptionist that books appointments via Google Calendar. Built with FastAPI + Claude API.

Playing to Win strategy framework as a Claude Code skill. Pressure-test business ideas through Roger Martin and A.G. Lafley's choice cascade.

Run an open-source repo as its maintainer: triage the tracker, delegate the work, review hard, merge on green. One loop, one version, and per-repo differences in config instead of prose.

Agent Kaizen skill for Claude Code & Codex: faithful, identifier-preserving reference for Adobe Creative Cloud, Acrobat, Photoshop, Illustrator, InDesign, InCopy, Premiere Pro, After Effects, Audition, Media Encoder, Firefly. UXP, ExtendScript, CEP, batchPlay, plugin SDKs & REST APIs. Routes to 11 product subskills.

Multi-provider API gateway for routing requests to any LLM — manage providers, aliases, and monitor requests from a web dashboard.

Setup your RAG agent using FastAPI with this ONE package!

An autonomous AI Operating System for coding, automation, creative generation, and intelligent task execution.

你的私人摄影指导团队

🌍 Agentic geopolitical risk analyzer — discovers second-order asset exposures through historical evidence, not LLM hallucination. Agent proposes, deterministic layer verifies. LangGraph + ChromaDB + FastAPI.

Tracify is an AI observability platform that helps developers monitor, debug, and optimize AI agents in production. Trace every decision, tool call, failure, retry, latency spike, and cost in one clear view.

Agentic habits: standing behaviour for a coding agent in three tiers. Stated rules where the trigger fires, a Stop-hook gate for what must not fail, and a read-only judge that scores adherence from evidence.

Dark Linear/AI-OS style technical diagram skill — cyan/purple/lilac duotone SVG diagrams

🔍 A beautiful web viewer for AI agent session files. Browse Claude Code & OpenClaw conversations with chat-style UI, timeline visualization, and zero setup.

Pi coding agent 的桌面工作台：Web UI + Electron 壳，内置右侧浏览器控制、Univer 表格/文档/PPT 深度集成、会话可视化与团队协作

Your AI agent writes code — but does it know WHY your code exists? Decision provenance, quality contracts, and AI anti-pattern detection for coding agents. 19 languages. MCP compatible.

👀 给 AI agent 一双能看见运行时的眼睛 — Vite 自愈遥测插件（API/错误/代理 header 三类结构化日志 + 本地 cookie 修复），让 agent 自我诊断·修复·验证。npm: vite-plugin-agent-eyes

Embeddable agent runtime and workbench for teams that need sandboxed execution, host integration, platform governance, and reusable AI infrastructure.

AI-powered SEO article generator with agent-based pipeline — SERP analysis, editorial briefs, hybrid quality scoring, multi-provider consensus (Claude + Gemini), and automated edit loops

A Claude skill and standalone CLI for inspecting, debugging, and injecting code into a running Steam client (desktop or Steam Deck) via the Chrome DevTools Protocol (CDP)

An open-source AI agent that tests your investment thesis (why you own a stock) against new company filings—with exact citations, human review, and version history. Built for investors who care about the business behind the stock—whatever their investing style.

⚡ Structural code intelligence & call-graph engine for C# and .NET — built for AI coding agents under hard token budgets.

Drop-in operating systems for knowledge workers. Product Manager OS: 53 skills scored 100/100 in CI, an operating brain, a memory that learns your product, and a loop that builds you new skills from the work you repeat. Runs on Claude Code, Cowork, Codex, Cursor, and Gemini CLI.

Codex、Hermes 与 Claude Code 的本地桌面协作中枢

One local dashboard for many Git repos - status, Docker, and your Codex/Claude/Gemini CLI sessions.

Persistent knowledge and validation-gated skill evolution for coding agents, based on arXiv:2608.27454.

Practical patterns, starters & CLI tools for memory engineering with AI coding agents. Stop re-explaining the repo — design the memory. Get a Memory Ready score.

Markdown templates generator for agentic-assisted coding

Use the official Grok Build CLI as a controlled external subagent for Codex via ACP and MCP.

Drop-in .claude/ config that turns Claude Code into an autonomous academic paper production system. 6 agents, 47+ skills, 5-phase pipeline from literature search to submission-ready manuscript.

Free, open-source MCP server for Roblox Studio. 31 tools, a console panel that runs coding agents, editor-safe script edits.

🤖 Explore essential AI engineering resources, from research papers to tools, to help software engineers transition into AI roles effectively.

A Claude Code skill that checks a branch the way a careful reviewer would, then pushes it. Two adversarial model reviews, a blocking security pass, a coverage gate, and a live run. No dependencies.

Stealth browser automation for AI agents — unattended, headless, N parallel persistent logged-in Chromes. One MCP server + CLI, credential vault with TOTP/IMAP 2FA, vision clicking, prompt-injection scanning. Self-hosted, real Chrome. 1,061 tests.

Companion tools and reference patterns for SPECLAN, the VS Code extension for spec-driven development with hierarchical Goal/Feature/Requirement specs as Markdown files in Git.

Bipa CLI — payment rails for AI agents. Pre-built binaries and managed installer.

Turn recorded screen activity into deployable n8n workflows.

Let AI agents QA your macOS app in the background. No synthetic input, no stolen focus.

Open-source, self-hostable system of record for team decisions capture the why, review the outcome.

Cross-tool persistent memory for AI agents — local, private, no API key required

13 negotiation & persuasion skills for AI agents, distilled from 36 books: CIA psyops, FBI behavior files, propaganda science, dark psychology. Plain English. MIT.

Content-addressed SKI compute core with bit-exact consensus. Spec-first, multi-model reviewed, machine-verifiable conformance.

Decision boundaries and installable skills for AI coding agents.

Work anywhere. Orchestrate coding agents and real terminal workflows from one Web command panel.

Multi-provider AI coding assistant with transparent vision proxy, usage metrics, and OpenAI-compatible local gateway.

Local action firewall for AI coding agents: scans what the agent reads, taints the session, blocks dangerous follow-up actions

Scheduled tasks for your coding agents, in the terminal. A prompt, a cron line, and a fresh git worktree per run — on Herdr. One YAML file, no store, prebuilt binaries, per-automation model, catch-up after sleep, and a live board.

Terminal-based AI coding agent — provider-agnostic, MCP-ready, with live reasoning streams, prompt caching, plans mode and a skills system. Pure ES modules, zero build step.

Local AI automation workbench for Siemens PLC and industrial-robot R&D: MCP, TIA Portal V21, PLCSIM, guarded writes, and offline-first verification.

Your coding agent makes the demo video — a narrated, captioned MP4 from one storyboard.json. Deterministic replay re-renders it in CI (no key, ~$0). Open-source (MIT) alternative to Screen Studio / Clueso / Demosmith.

Watch your Herdr coding agents as an 8-bit town. Read and answer them without leaving it.

Private, local-first continuity for terminals and AI coding agents—from desktop to phone, with a self-hosted encrypted PWA.

Measure a repository as a place for a coding agent to work, and lay the foundation where it has none. A Claude Code plugin: hooks, guards and gates proven to turn red, layered docs, a repo graph, and a five-dimension assessment you can take again later.