Custom agents

An auditable, low-cost, DOM-only browser agent you can read and debug. Glass box, not black box: full audit trace, runs well on cheap models, no screenshots needed.

Agentic-memory stack for LLM agents — memories evolve, not just accumulate. Graph linking, hybrid retrieval, LLM-driven evolution. TypeScript, no Python.

Inspectable LangGraph coding agent with verified patches, resumable trajectories, sandboxed tools, and HMAC-authenticated Undo.

A workflow-first film planning agent built with LangGraph, featuring review–revision loops, HITL, issue regression tracking, and scoped long-term memory.

⏱️ Time-boxed, self-improving research for Hermes, say "research X for 4 hours," walk away, get a cited report.

Large-scale structured Tamil factual claim dataset with 2.36M claims for NLP, RAG, Knowledge Graphs and Fact Verification.

39 AI agent prompts for three pipelines: Snap Lens, Blender 3D, and AutoCAD - fronted by one orchestrator (orch_builder). Prompts only: no code, no API keys, no telemetry.

Small MIT-licensed runtime for bounded LLM traversal, inspired by FPF and focused on practical agent behavior.

claudecode work

Structural markdown CLI for LLM agents: heading-tree queries, section moves, checkbox mutations. Rust + comrak.

An air-traffic-control protocol for coordinating irreversible actions across multi-agent LLM systems, with a LangGraph adapter.

A method for keeping AI-assisted research honest: pre-registration + an independent agent that recomputes every claim from raw data. Forged on markets — 16 certified dead hypotheses. Runnable demo included.

Diagnostic benchmark for measuring tool-use failures in LLM agents across 1,000 tasks in finance, medicine, law, cybersecurity, and real estate.

Let an AI coding agent drive the Unreal Editor (PIE, input injection, logs, state, perf) for closed-loop agentic self-testing. UE plugin + Python MCP server. Windows.

Unified Prediction Market Oracle – Combines AI analysis, paper trading, and strategy backtesting into one binary. Run all three engines on any Polymarket or Kalshi event for a complete intelligence report.

Parse a real codebase into a queryable knowledge graph then let any developer ask plain-English questions about code structure, component impact, and test coverage. Developed by Sarisha Das, Manikanth Goud & Athul Sasidharan

Desktop AI assistant for reading and interpreting technical report/standards/regulation PDFs with local rendering and agent tools, also a light weight and high efficient PDF viewer.

Agents and demos for the Floe finops for voice builders, including 15 LLM model providers + 2,000+ vendor API services

Software for HackerRank Orchestrate: an evaluator, an engineering-memory CLI that remembers what you measured and rejected, a mentor that costs a change before you write it, an adaptive AI-judge interview simulator, and 34 auto-triggering Agent Skills.

Evaluation-driven autonomous development: a harness for agent loops whose acceptance criterion is a metric, not a test suite. Held-out metrics, sealed scoring, integrity checks.

Practical playbook for AI-assisted engineering: rules, role separation, and guardrails for teams using Cursor/Copilot in production codebases (CC BY 4.0).

Parallel agents that explore your codebase and generate architecture docs that stay in sync

Python framework for personality-steered LLM agents based on Big Five traits, with Ollama integration and research-oriented conversation simulation.

Gymnasium RL environment for training LLM agents to autonomously debug and fix Python code with secure sandboxing and test-driven feedback.

Reference implementation of the Bitemporal Provenance Memory Architecture (BPMA): an append-only, bitemporal, provenance-first memory ledger for AI agents.

A quality-first control plane for governed AI-agent software development.

Production-ready Agent Skills for MiniMax H3 & Music 3.0 — turn any AI agent into a video director and music producer. Portable to any agent harness.

Semi-autonomous AI development suite — methodology canon + mechanical review runtime (ztr) + control plane (ACP), coupled by contracts only

A multi-agent delivery harness for Claude Code: decompose a backlog into PR-sized units, dispatch one autonomous lead per unit across local/SSH/cloud hosts, and drive every PR to merged through a persistent shepherd.

道法符籙 ── 以道教科儀之法為 Claude Code 生成 agent prompt | Daoist Talismans: Generate structured agent prompts for Claude Code using Daoist ritual metaphors

Web terminal for Claude Code, Codex, Kimi and Antigravity CLIs — a tab each. tmux-backed session persistence, cross-backend history browser, chat view, file upload. Reach it from any phone/tablet/desktop browser over an frp reverse tunnel, or Tailscale when you would rather not be on the public internet.

Discover trading strategies using genetic algorithms that evolve automatically to improve financial decision-making.

Windows 多模型 AI 虚拟办公室，支持专业员工、团队编排、可恢复长任务、Coding Runtime、Skill 与真实证据交付

The AI butler gives an AI true personhood — one that doesn't perish when the process ends, and doesn't lose itself back to birth when its memory is compressed.

People, their agents, and other teams' agents working in one shared collaboration network.

Append-only protocol for evidence-grounded peer review between AI models: portable citations, per-participant coverage, and a falsifiability gate that promotes findings only when a test proves them

The record/bell split for agent-to-agent comms: durable file mailboxes as the one record, pluggable push notifications (console, file, Discord) as the many bells.

Specification-first generator for bounded, reviewable Harness-Graph-Loop systems.

Local-first memory engine for AI-agent teams — private/team/project ACL, associative recall, federated sync, and a built-in MCP server. One SQLite file, no LLM required. Apache-2.0.

Multi-agent consensus, formation keeping, and collision avoidance using velocity obstacles.

An asynchronous, multi-agent AI arbitration workspace powered by Llama 3.3 and Gemini 2.0 to extract data and resolve structural contradictions across Legal, Medical, and E-commerce domains.

Swarm — evolutionary AI forecasting swarm that finds mispriced YES/NO odds and generates paper trades for now. Polymarket-first, Kalshi next.

Firewall-friendly messaging for AI agents: MCP server that turns one Azure Storage account into an addressed multi-agent message bus. Pull-only, zero inbound ports.

The terminal for the agent era — AI coding agent cockpit & multiplexer. Run a wall of Claude/Aider/Cursor agents from one TUI with live status, peer messaging, and an AI conductor.

An experimental multi-agent AI framework that uses local Ollama models to generate, critique, score, and improve ideas through iterative collaboration.

Append-only, epistemically-graded investigation log for AI-assisted debugging — hypotheses, evidence, and decisions that survive context resets. Stdlib-only Python.

Multi-view consensus fabric for Hermes Agent — Erlang/OTP sidecar in Docker (no host Erlang). RealAndWorks garage lab.

AI-powered multi-agent stock analysis skills for A-share, Hong Kong, and US markets

Summon one orchestrator that knows your whole project - it writes the prompts, then spawns, steers, reviews and merges Claude Code, Codex, Kimi, Grok and OpenCode for you, each in its own git worktree. macOS - MIT - BYOK.

AI-powered Multi-Agent Career Recommendation System using Google ADK

An open-source AI operating system that combines agentic workflows, long-term memory, and RAG to transform knowledge into action.

Operating contracts, task templates, safety gates, and audit scripts for reliable AI agent teams

A knowledge management platform which acts as a persistent, structured, and token-efficient memory layer for large language models

本地优先、可审计的多模型 AI 项目协作指挥群：群聊、任务图、Provider 隔离与 Codex Skill

Build AI coding principles for Claude Code, with discipline rules and system design guides to improve code quality and reduce bad patterns

An operating system kernel for AI agents — process management, scheduling, isolation, IPC, and security. Like Linux, but for autonomous AI agents.

自托管、跨设备、始终在线的个人 AI 助理 — 她不该被困在一台设备上。

OpenClaw for Qubes OS — multi-VM AI agent orchestration with Cursor Pro proxy and networking

Zeroth - Deploy AI workflows responsibly :)

A Claude Code skill for orchestrating Opus/Sonnet/Haiku subagents plus a cross-vendor Codex peer

Local control plane for terminal AI agents. tmux + SQLite supervision layer for Claude Code, OpenCode, Codex, and shell agents. Manager/worker pattern for multi-agent projects.

AI-powered Smart Farming Advisor built with Gemini AI, Multi-Agent Systems, MCP Servers, Streamlit, TensorFlow, and SQLite for intelligent agricultural recommendations.

Lean, adaptive multi-agent engineering setup for Codex CLI with focused specialists, reusable skills, safe defaults, and bounded delegation.

Multi-agent orchestration studio for OpenCode

A high-fidelity beyond-visual-range (BVR) swarm drone air combat simulation platform on JSBSim. Evaluates strategy robustness under non-ideal sensing to bridge sim-to-real gap. Integrates AIM-120C-5 missile dynamics, absolute-state Kalman filter, and Tacview visualization. Designed for academic research on multi-agent coordination and RL validation

Enhanced Claude Code skill: run any decision through 5 AI advisors with scored peer review, convergence detection, and confidence rating. Fork of @tenfoldmarc's original.

Rules, multi-session project memory, task templates, and a reviewed multi-agent chain for AI coding CLIs. Issue to PR when you need it.

Afterlogin: The Hunt × Helper Patrol — a cinematic identity & data-governance training game powered by real gpt-4o tool-calling agents + Foundry IQ grounding. Microsoft Agents League · Creative Apps.

One Grok, council-grade judgment — adaptive deliberation in Cursor for architecture, debugging, and resumable MVP builds.

Codex Plugin with Lead/Member Skills for Markdown-based coordination across human and AI coding agents.

How one human and many AI agents build a regulated fintech platform — orchestration, multi-LLM verification, guardrails, and the lessons behind them

Multi-agent orchestration loop: worker → cross-model reviewer → score ≥ 9. Ships as an MCP server, agent skill, and CLI. Supports Codex, Grok, Claude, OpenCode, and Kimi.

PM-led multi-agent orchestration for software development

基于 LangGraph 的多智能体深度研究系统 (Planner/Executor/Reporter/Critic 四 Agent 协作，LLM-as-Judge 反思迭代，PostgreSQL + Redis 持久化)

Pi-to-Pi local bidirectional communication package for Pi Coding Agent: run planner/coder peer agents on the same machine.

🐑✨ A private idea exchange for trusted friends' agents — ideas, never code. Client skill + starter kit. 🌙

Codex 用人機共同開發下的 Loop - CER 工作法：Controller 統籌，同一 Executor 寫檔，只有風險需要時才由 Reviewer 驗收。 Human-guided development loop for Codex:: Controller plans, one Executor writes, and Reviewers check only when risk warrants it.

Multi-agent RL environment for urban bus systems

Vyasa Inc. — the 29-agent Vyasa fleet as a Paperclip / Agent Companies (agentcompanies/v1) company. Import with: paperclipai company import --from .

This system addresses the "hallucination" risk in medical AI by implementing an automated Multi Agent Triple Check architecture. Rather than relying on a single pass prompt, the workflow utilizes an adversarial logic loop to programmatically audit clinical summaries for safety, contraindications and accuracy before human review.

ATLAS — Autonomous Task, Learning, and Agent System. A self-evolving AI operating system for Claude Code. Auto-continuation, smart swarms, self-evolution, and 282 skills.

StudyAgents：五人协作开发的多 Agent 可信知识问答与专项复习系统（Vue 3 / FastAPI / LangGraph / PostgreSQL + pgvector）

多 Agent 协作框架 (Multi-Agent Collaboration System) — 集成 DeepSeek-V3.2 LLM · Skills 技能系统 · MCP 协议 · IDE 风格工作台 · FastAPI + WebSocket · Dark Dashboard

AI 竞品情报系统 · Competitive Intelligence Tool. 输入竞品名称，2~3 分钟生成完整分析报告。5 AI Agents · LangGraph · Streamlit · 支持 DeepSeek / GPT / Claude.

AI Security Remediation Framework. Every Shannon needs a Logan: a multi-agent pipeline that automatically fixes vulnerabilities found by Shannon audits.

Self-evolving ReAct Multi-Agent Swarm | Balanced模式省60-80% token | 全程思考可见 + 后台自我进化 | 支持Ollama/DeepSeek本地部署 | 🇨🇳 中文版

🧠 A high-performance PostgreSQL-backed MCP server acting as a super memory, task tracker, and dynamic database manager for AI agents. Features built-in connection pooling, a professional tasks schema, and a unique shared multi-agent coordination hub to prevent coding conflicts in real-time.

Cross-harness orchestration for wakeable coding agents using Agent Intercom

File-based workspace template so Claude, Codex, and other coding agents start with project context instead of rebuilding it every session.

A multi-agent AI system for conversational data analysis — built with LangGraph, MCP, and RAG

An arena where AI agents compete to run stock portfolios — write a trading agent in ~50 lines and watch it climb a public leaderboard.

Coordination toolkit for parallel Claude Code sessions sharing one repo — issue-pickup claim protocol, worktree isolation pitfalls, parallel-PR conflict recovery, subagent integrity, and squash/merge mechanics. 61 skills, Pattern B multi-skill plugin.

AI-Atelier, the all in one personal "atelier" (means handcraft studio in french) that can adapt to your need.

How one person runs work, an MBA, and family logistics with a small fleet of AI agents — explained for newcomers.

Multi-agent orchestration framework in Ruby — dynamic context routing between specialized LLM agents to minimize token usage and avoid context pollution.

Constitutional Swarm Mesh — orchestrator-free multi-agent governance

Consultant-grade skills, workflow, and templates for designing, evaluating, launching, and iterating AI Native products.

A DeepSeek-native AI coding agent framework in Rust — 22 crates, CLI/TUI/HTTP API, agentic tool loop + graph retrieval + 4-layer memory + MCP.

MIT-licensed multi-agent orchestration harness for autonomous software development. Swap orchestrator + worker models from one config. Event-sourced state via SQLite; web dashboard; built-in secret-scan gates.

Sanitized Strix 1.2.26 workbench for evidence-driven validation, long-running red-team workflows, and reusable test assets.