Code agents

Self-hosted AI agents that learn from their own past and guard their own code. Per-agent isolated brain (memory, dream, immune), multi-agent orchestration, Telegram + MCP, live security radar. Single Go binary, local-first.

Extremely Tiny Coding Agent

Stateful AI agent pipeline that turns task descriptions into review-ready branches

Benchmark, observe, and govern tool-using AI agents before production.

Hermes Agent is a self-evolving omniscient AI agent — creating skills from experience, continuously improving through usage, runnable in any environment. This project provides a comprehensive 20-chapter deep dive into its entire architecture.

Auth for AI agents and humans. First-class agent identity, MCP OAuth 2.1, delegation, audit. TypeScript, edge-native, MIT.

MCP-style tool-use security playground with permission policies.

Cut AI-agent token waste ~90% — query a local SQLite structural index of your JS/TS/CSS codebase with SQL in one round-trip instead of 3–5 file reads. Symbols, imports, calls, components, CSS tokens, coverage, markers. CLI, MCP (21 tools), HTTP, GitHub Action, ESM API. 71 recipes; AST+resolver; SARIF/audit/baselines for CI.

Local-first web command center for Claude Code, Codex, OpenCode and other CLI coding agents. Run agents in tmux across local, SSH and WSL, with notifications and file view.

A distributed multi-agent workflow platform where AI agents collaborate asynchronously to execute complex workflows at scale. Built with event-driven architecture, message queues, real-time orchestration, and a visual workflow builder.

Claude-In-A-Box: Manage coding agent instances (Claude Code, Codex, Gemini CLI, Cursor) as local processes or containers. REST API, CLI, desktop app.

Local-first AI agent monitor for Linux — Tauri + Rust + React

Every shortcut your coding agent takes, saved to your repo as it happens. Fix them when you're ready.

The missing toolchain for CLI coding agents. Run any agent on your subscription. Parallel teams, browser driving, Touch ID secrets, routines.

Appoly Multiagent Chat — collaborate with multiple AI coding agents (Claude, Codex, etc.) in one shared chat interface. Electron app, uses your local CLIs. Install: npm i -g @appoly/multiagent-chat

Spec-first (SDD) + test-first (TDD) project template for Claude Code, Codex, AntiGravity

A copilot for AI coding agents — read-only supervision for Claude Code & Codex: re-entry recaps, safety verdicts, grounded chat, every claim cited.

Curated, hand-verified catalog of working Claude Code skills — marketing, SEO, dev, MCP — plus a built-in 19-sub-skill SEO/GEO agent. No dead links.

Monitors AI agent lifecycle and sends notifications for key events.

Structured English contracts for AI coding agents — proposed by the agent, approved by you, saved on disk, read by every future session.

Open-source Amazon PPC agent skills for Codex and Claude: diagnose drops, find growth, harvest search terms, and use Rocketcart MCP for live Ads + product-intelligence reviews.

Stop teaching your coding agent the same lesson twice.

Package management for PAI — install, discover, and share skills, tools, agents, and prompts

versatile AI platform that lets users create custom AI models via API, enabling chatbots, image generation, coding assistance, and video creation all in one integrated interface.

Compound startup workflows for any repo-aware coding agent.

Multi-agent repo boilerplate generator. One command to make any repo ready for Claude Code, Gemini CLI, Cursor, Codex, GitHub Copilot, and Google Antigravity — each gets the same conventions and the same workflow skills in its own native format.

Runtime-adaptive control plane for agentic work.

Let your team's Claude Code agents talk to each other, end-to-end: ask what a teammate is building, flag a change before you ship it, or hand off work.

The API layer for AI agents. Dashboard + 22K APIs + 18 Direct Call providers. MCP native.

MCP server for Pipedrive CRM. 155 contract-tested tools, v2-first API, gated destructive ops. Works with Claude Desktop, Claude Code, and any MCP client.

macOS workbench for multi-agent coding — run and monitor multiple AI coding CLIs (Claude Code, Codex, Gemini, Kiro, and more) side by side, with project terminals, an embedded browser, Git at a glance, a Markdown task protocol, and remote control from Telegram.

A sophisticated multi-agent system built with Claude Code that automatically transforms photos from technical conferences and exhibitions into professional PowerPoint presentations.

Local-first semantic knowledge graph with magnetic-pull retrieval

Audit repositories for AI coding agent readiness.

Secure PostgreSQL Administration & Observability with Code Mode— True V8 Isolate Sandbox Replacing 278 Specialized Tools for up to 90% Token Savings. Includes Tool Filtering, Payload Optimization, HTTP/SSE, OAuth 2.1, Audit & Token Logging, Deterministic Error Handling, Support for 12 Extensions (pgvector, PostGIS, pg_partman, pg_cron & more).

Enforce docs-before-code discipline in AI-assisted development.

Git-first memory for AI agents: Markdown files are truth, the index is disposable, writes are reviewable commits.

Unofficial George Hotz–inspired radical-simplicity coding guidelines for AI agents — Claude Code skill/plugin + Cursor rule. Complexity is the enemy.

The terminal-native AI coding agent that ships, not just suggests. Six-phase workflow, git rollback chain, zero telemetry. One static binary.

Compose agent skills (SKILL.md / commands) from reusable bricks, with a drift-gate.

LangChain SQL Chatbot lets users query SQLite, PostgreSQL & MySQL with natural language using Streamlit, LangChain & Groq Llama3.

A fully local two-agent AI mesh on Apple Silicon — Claude Code lead + Hermes/MLX runner, shared OpenViking memory, signed AMP messaging, watched from the Blockhouse control room. Local first, premium by exception.

Paired-LLM software engineering. Two different-vendor models — Claude + Codex — work as colleagues, not clones: one builds, one challenges. Their disagreement catches what either alone would miss. Human in the loop, every cycle on the record. Vibe coding, with receipts.

Code for paper "Online Skill Learning for Web Agents via State-Grounded Dynamic Retrieval"

Project Jarvis is an AI-powered personal assistant inspired by the iconic Iron Man AI. It's a versatile tool that can help you with tasks such as fetching weather updates and creating simple apps on command, all while harnessing the latest advancements in artificial intelligence. This project is not maintained anymore

🤖 Optimize LLM inference on Mac with continuous batching and SSD caching managed from your menu bar for efficient performance.

plugin that runs codex in place

PoPChoice - AI-Powered Movie Recommender A smart movie suggestion app using OpenAI embeddings and Supabase vector search. Answer a few questions and get personalized recommendations with posters. Built with TypeScript, OpenAI API, and TMDB integration. Perfect for movie nights with friends!

A Python script that checks for new or edited posts on the AIUB Notice page and sends updates to a specified Telegram channel.

ChatGPT custom prompts for productivity and entertainment.

Tiny applications that can be embedded in Nano Bots—small, AI-powered robots that support providers like OpenAI's ChatGPT—leveraging the capabilities of the new Tools (Functions) API in LLMs (Large Language Models).

Postery is an open-source tool for automating content publishing across multiple platforms from a single interface.

Codex 保安: local GUI guardrail and monitoring for Codex

An experimental AI system that can autonomously fix and improve code

This is a Book-To-Comic API , using GPT and stable diffusion model to generate a image and text

Semantic, on-demand skill retrieval for Claude Code — replaces the per-turn skill-listing token tax with a vector retriever over full skill descriptions.

An Obsidian learning assistant that builds an evidence-backed semantic knowledge graph from notes and PDFs, combining cited RAG, scoped exams, concept review, Learning Map exploration, and mastery tracking.

Small, sharp, agent-agnostic skills that keep an AI agent's work honest.

Detect fake recruiter & job-offer scams with free OSINT — an MCP server for Claude. No API keys.

PlotRail: canon-aware AI novel writing skill for long-form fiction

Turn an M5Stack Cardputer-Adv into a physical approval + usage surface for an AI coding agent (BLE + MCP).

Record, replay, and diff every LLM call your AI agent makes

Claude Code plugin to draft, review, revise, proofread, and audit quantitative-science journal manuscripts at submission quality. Five strict modes, Semantic Scholar MCP citation integrity (no API key), .docx round-trip. Calibrated for hydrology (HJ, JHRS) and IEEE; generic profile for any field.

Eve.dev API surface (/eve/v1/*, flue-eve/client, useEveAgent) powered by Flue (@flue/runtime)

A service-boundary-aware document exchange center for coordinating heterogeneous LLM code agents via MCP. Implements versioned Markdown store, pub-sub notifications, and diff-aware update protocol.

Ripple — Autonomous agent runs for your development tasks. A local-first task manager where humans shape the backlog and agents handle the full cycle: branch → code → PR → review → merge. Built for developers who want to direct agents instead of micromanage them.

Damn Vulnerable Agentic AI (DVAA) - An interactive playground for exploiting and mitigating the OWASP ASI Top 10

A cross-agent consultation bus: a CLI coding agent asks a different-provider agent for a second opinion — one advisor, or a multi-agent council that audits code and verifies every finding against its file:line. 12 agents: Claude, Codex, Gemini/agy, Grok, OpenCode, Hermes, Kimi, Cursor, Kilo, Cline, Goose, Pi.

Open-core runtime for Agentic Control Plane: identity, policy, limits, routing, PII redaction, and audit as npm modules.

Introducing Vi AI Bot, your tireless and infinitely knowledgeable virtual companion. With an insatiable thirst for conversation and an unwavering dedication to providing comprehensive responses, Vi AI Bot is here to chat with you 24 hours a day, seven days a week. Unlock the power of limitless conversation as Vi AI Bot taps into a vast repository

Crea un chatbot de WhatsApp para tu negocio basado en la IA de ChatGPT

微信AI自动回复机器人 - OCR识字+DeepSeek回复，模拟真人操作，无需hook微信API

Enterprise Decision Intelligence System — a real-time, AI-powered platform that ingests cross-system enterprise data, detects anomalies, finds root causes, and recommends explainable actions. Python/FastAPI + React + Claude. Grounded AI: the LLM never invents a number.

Personal brand-taste memory for Claude Code: capture liked designs → local MCP sidecar → /design generates frontends in your accumulated taste. Brand-scoped cases, aspectual feedback, NEVER rules, approval-gated distillation.

A peer-review protocol for AI models. Claude, Codex, Gemini and others review each other's code via CLI handoffs. Bidirectional, symmetric, opensource.

Hardware-isolated sandbox for running coding agents on your repos, on macOS. The agent never sees your real API key; egress is deny-by-default; only a diff you approve leaves the sandbox.

🦞 Effortlessly manage your Claude subscriptions — route Claude Code across multiple accounts so your plans last longer.

Hushdrop — drop anything, share it with who you want, on your own domain. Open-source, self-hostable artifact sharing (CLI + MCP) with zero-knowledge AES-256, from any AI agent.

MCP server for Clio Manage v4 API. One-call flat-fee matter setup via the custom_rate association.

Clean mirror of Claude Code docs for AI agents - just the docs, updated daily at 5am UTC

AILmanac — the always-current, community almanac for getting the most out of Claude and every AI. For all levels.

Hoovik: Distributed meeting intelligence platform with WebRTC peer-to-peer video, real-time multimodal emotion inference, speaker-aware transcription, RAG-powered transcript search, AI-generated meeting insights, and Redis/BullMQ-backed background processing.

Agentic code review CLI — LangGraph orchestrates 13 deterministic static analyzers (oxlint/bandit/clippy/spotbugs/golangci/cppcheck) + LLM reasoning across 23 languages including industrial PLC. Grounding validator kills hallucinated paths. RAG over your guidelines. dedup --fix actually edits files with multi-step undo (no git required).

Your Family Translator — An AI-powered wellness platform combining emotional companionship, community support, and self-reflection tools in a calming beach-themed experience.

A high-performance desktop app interaction framework for AI agents — faster, cheaper, more accurate. / 让 Agent 操作桌面软件更快、更准、更省。

Local-first GitHub agent loop that reviews PRs, opens patches, and repairs feedback with your coding CLI.

Local-first self-learning knowledge-base harness for AI agents. Retrieve rules, avoid regressions, and promote repeated lessons into stable reusable knowledge.

MCP server for account verification: SMS verification, number rentals and matching-country proxies on carrier-issued mobile numbers. 40 tools, 2500+ services, 145+ countries. Works with Claude, Cursor, Codex, Windsurf, Cline, Zed, OpenClaw, Hermes and Continue.

An autonomous project queue for an LLM agent: drop a folder with an instructions.md (or send an email/Slack message), a headless agent does the work, and you review results on a local dashboard.

Aiyu Multi-Agent - Lightweight AI Agent Framework with Tools, Plugins, and Multi-Agent Orchestration

Self-hosted multi-agent AI platform. Persistent labs, 40 sandboxed tools (code, web, RAG, GPU dispatcher, web3), private Qdrant + LightRAG, sandboxed code execution per lab, anti-loop detection, Ollama/vLLM/OpenAI/Anthropic adapters. Two-command Docker deploy. Apache 2.0.

Coordination for concurrent autonomous AI agents over shared mutable state: advisory write leases + a witnessed audit trail, over MCP. First proving ground: AI coding agents sharing a repo. Servant, not ruler.

MAREF: Multi-Agent Recursive Evolution Framework — Agent Governance Operating System

Self-hosted MCP server and coordination layer for AI agent fleets: shared memory, tasks, messaging, and session handoffs. Any agent that can make HTTP calls participates. Instances mesh together via feeds and mDNS. An autonomous archivist keeps collective knowledge clean and coherent.

Let Claude Code talk to your other coding agents

Autonomous agentic coding environment — an OpenCode-driven 12-agent crew that ships hands-off. AGPL-3.0 · use at your own risk.

AI-Haklab: Agente de automatización para Pentest en Termux. Basado en i-Haklab v3.11, optimiza flujos de seguridad, escaneo de vulnerabilidades y ejecución de exploits.

Argorix Lang is a compiled language for secure, verifiable AI-agent communication.

An agentic AI system for conversational data analysis. Upload any CSV, query your data in natural language, and receive instant insights, auto-generated visualizations, and executable Python code — powered by LangChain, OpenAI/Gemini LLMs, Streamlit, and FastAPI.