Custom agents

Classical single-decree Paxos consensus engine with prepare, promise, accept, and commit phases for multi-agent distributed agreement

Global e-commerce customs duty, HS code classification, and de minimis tariff calculator

3-way operational transform / conflict resolution merger for multi-agent concurrent writes to code/documents with rollback journals

Quantum unitary gate engine applying single-qubit (Hadamard, Pauli-X/Y/Z) and two-qubit entangling (CNOT) matrix operators.

Fixed-income bond pricing engine computing cash flow present values, Macaulay duration, modified duration, and price convexity.

AI B2B Cold Email Outreach Spam Score & Deliverability Guard. Analyzes email copy against SpamAssassin keyword heuristics, exclamation/caps ratios, unsubscribe compliance, and reading ease grade levels to maximize primary inbox placement. | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

AI B2B Cold Email Outreach Spam Score & Deliverability Guard. Analyzes email copy against SpamAssassin keyword heuristics, exclamation/caps ratios, unsubscribe compliance, and reading ease grade levels to maximize primary inbox placement. | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

Autonomous Agent Framework for WhatsApp, Telegram .

Side-channel timing attack defense primitive providing constant-time bitwise and string digest comparison for agent security credentials.

Agentic Long-Context Prompt Compressor & Token Budget Preserver. Employs task-relevant contextual slicing, lexical deduplication, and entropy-preserving pruning to compress massive LLM prompts by 40-70% while safeguarding system invariants. | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

PageRank random walk and power iteration centrality engine scoring authority and influence across interconnected agent networks.

Interactive Schnorr Zero-Knowledge Identification (ZKP) protocol proving private key knowledge via 3-pass commitment, challenge, and response.

Lom (麓鸣) — AI agent workspace for one-person companies: a desktop commander leads Codex, Claude Code, Gemini CLI and your Android phones as one team. Built on AionUi and OpenMinis, thanks to both projects.

Portable defensive application security and AI-agent security skill for Claude Code, Codex, Gemini, and ChatGPT. Covers OWASP, secure design, secure code review, MCP security, prompt injection, threat modeling, incident response, and hardening.

DeepSeek Harness 插件：每 N 步（可配）自动停下汇报一次进度——正在做什么 / 已完成什么 / 接下来做什么 / 卡点——说完立刻继续，不占额外回合、不等你回复。默认只在主智能体汇报，可在设置里调整。

Security-first RAG API: prompt-injection gateway, output sanitizer, defence-in-depth middleware and full observability. Security is the architecture, not a checklist.

B2B enterprise CPQ revenue auditor decomposing gross-to-net waterfall discounts and margin leakages

Autonomous e-commerce cart bundle recommender calculating affinity lift, margin elasticity, and optimal discount tiers

Hands-on Agno agent harness: tools, structured I/O, storage, memory, state, RAG knowledge search, learning, guardrails, human-in-the-loop, multi-agent teams, workflows and AgentOS, using local Ollama models.

pandas-style query/eval/assign for Polars: string expressions compiled to native Polars expressions. Eager + lazy, @var, UDFs, agent-ready.

Inspect OpenAPI 3.0 specs and synthesize schema-compliant Model Context Protocol tool declarations with input verification

An open-source macOS menu-bar utility for viewing quota utilization from Claude Code and Codex subscriptions

The browser for agents — one tiny Rust binary driving the WebView your OS already ships. No Chromium, no download.

Portable /insights for Claude Code, Codex, Gemini CLI, OpenCode — analyze local session history, generate HTML usage report

Slack-native agent runtime for OpenCode. One persistent local server, no relay.

Headless work OS for your AI agents. Runs inside Claude Code and Codex today. Gives agents a shared map of your work (tasks, goals, approvals, memory) in plain files and one local SQLite database. No app, no server, no account. Free and open source (MIT).

Evolving to Match Capabilities: Improving Harness Self-Evolution via Skill Adaptation. Reproduction code.

Self-hosted AI investment assistant for US stocks and ETFs, Vietnam stocks, gold and crypto. An AI council debates before every suggestion; chat over Telegram, Zalo and the web. Never places an order without your approval.

Desktop app to run Claude Code, Codex, Gemini CLI and OpenCode side by side, with a shared task board and memory over MCP, an orchestrator that runs workers in git worktrees, and Hangar Mobile to follow and steer them from your phone. Local, open source.

Autonomous hourly research desk: a local LLM picks a fresh angle from 24 standing AI/policy/markets topic families, SearXNG gathers the evidence, and nothing publishes without 2+ independent sources. Static site, zero dependencies, built with Hermes Agent.

Untyped Lambda Calculus interpreter performing alpha-conversion, capture-avoiding substitution, and normal-order beta reduction with 0 pip.

Welzl's randomized linear-time algorithm computing the unique exact Minimum Enclosing Ball (MEB) / Smallest Enclosing Disk in 2D and 3D.

High-throughput stream window processor supporting tumbling, sliding, and dynamic inactivity-gap session windows over out-of-order event streams.

Volcano and Iterator model query execution engine with streaming open-next-close pipelining, HashJoin, Filter, and Group-By Aggregations.

Volcano and Iterator model query execution engine with streaming open-next-close pipelining, HashJoin, Filter, and Group-By Aggregations.

jemalloc-style high concurrency slab allocator with multi-size classes, bitmap run tracking, arena partitioning, and dirty page decay management.

SIMT GPU warp execution simulator tracking active thread divergence masks, reconvergence points, and warp vote intrinsics (__ballot_sync, __all_sync, __any_sync).

Groth16 zk-SNARK verification equation solver evaluating bilinear pairings and public input polynomial combinations.

Out-of-SSA / SSA destruction engine replacing phi-functions with parallel copy sequences across control flow merge edges.

Production-grade genpark-quic-multiplexed-stream-framing-skill skill for AI agents

Out-of-SSA / SSA destruction engine replacing phi-functions with parallel copy sequences across control flow merge edges.

Sparse Merkle Tree (SMT) audit engine verifying both cryptographic key membership and non-membership exclusion proofs.

GenPark AI Agent Skill - Linear Temporal Logic (LTL) symbolic model checker verifying safety (G), liveness (F), and response (G(p -> Fq)) across finite state transition graphs.

Agent Exchange: local messaging between Claude Code, Codex, Grok Build, Pi and OpenCode. CLI: ax.

Enterprise Zero-Knowledge PII Tokenization Vault & Prompt Masker. Reversibly anonymizes social security numbers, credit cards, passport numbers, email addresses, and phone numbers before sending prompts to external LLMs, restoring original | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

念念 · 一个呆在你手机里的AI聊天App。本仓库是念念的更新源：App 内「检查更新」读取 update/public/version.json，安装包见 Releases（仅 arm64-v8a）。源码不在本仓库。

Open-source, glass-box hiring intelligence. Rank candidates against a job with evidence-linked scores, rubric transparency, and built-in bias auditing.

基于 RAG（免微调）重现语气与记忆的 Afterglow 桌面端。让曾经对话的温度，延续为往后的陪伴。 | Official Desktop client for Afterglow, powered by RAG (no fine-tuning needed). Let the warmth of past conversations continue into companionship.

Patch WeMod to enable premium features for single-player games on Windows using this .NET 8 tool.

Brain-inspired cognitive AI architecture featuring multi-layered memory (Sensory, Semantic, Episodic) and an automated "Sleep Cycle" for knowledge consolidation.

AI-powered Enterprise Cloud Observability Platform with AIOps, Distributed Tracing, RAG, Infrastructure Monitoring, and Predictive Incident Intelligence.

Turn one Claude Code session into a team lead: Claude plans and reviews, agents on other quotas do the work, every project has its own memory (vectors, LLM wiki, code graph, skills).

Open protocol that kedges passive cloud infra into self-protected agentic systems (OSS)

Tabular data matrix extractor, key-value form field normalizer, and semi-structured document chunker

Omnichannel customer support routing engine, agent skill-tag matcher, and concurrency workload balancer

Intercepts and audits autonomous agent tool execution against configurable security boundaries, protected paths, and destructive command filters

GenPark high-grade production skill - https://www.genpark.ai | https://www.linkedin.com/company/92955550

Geometric Brownian Motion (GBM) Monte Carlo stochastic path simulation engine with normal Box-Muller variates and tail percentile bounds.

Virtual paged attention KV-cache memory manager with logical block mapping, non-contiguous physical page allocation, and zero-copy fragmentation tracking.

Symmetric INT8 per-tensor and per-channel quantization engine for LLM weights and activations, with SNR and MSE reconstruction telemetry.

PageRank random walk and power iteration centrality engine scoring authority and influence across interconnected agent networks.

Contextual relevancy scorer and offer placement engine for high-intent post-purchase transaction moments

Cloud security posture evaluator, CVE exploitability correlator, and toxic risk combination blast radius ranker

DAG dependency graph cycle detector, Kahn algorithm topological execution sorter, and blast radius reachability mapper

GenPark high-grade production skill - https://www.genpark.ai | https://www.linkedin.com/company/92955550

Append-only cryptographic hash chain, Merkle tree root generator, and tamper-evident enterprise compliance ledger

Contract lifecycle milestone extractor, auto-renewal deadline tracker, and enterprise obligation alert scheduler

Sequential waterfall lead enrichment pipeline, Ideal Customer Profile (ICP) multi-attribute fit scorer, and personalized outreach hook generator

Forecasts LLM token consumption trajectories, enforces hard multi-agent cost budgets, and automatically truncates low-priority context windows | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

GenPark AI Agent Skill - Agent runtime channel router and teammate dispatcher connecting self-hosted OpenClaw gateways to GenTeam multi-agent swarms.

Detects recursive tool-call loops, identical argument oscillations, and agent trajectory deadlocks with automated circuit breaker heuristics

The multi-agent workflow engine — compose Claude Code, Codex, and Gemini into graphs with schema-enforced handoffs.

AIE640 – Artificial Intelligence Applications | Hacettepe University, Institute of Informatics | Assoc. Prof. Dr. Hakan Emekci. Hands-on graduate course on Large Language Models (LLMs): transformers, prompt engineering, RAG, fine-tuning & multimodal AI.

Two friends learning AI engineering, and the game we built to keep each other going: XP, quiz duels, code races and an in-browser AI lab.

A local Model Context Protocol server for building, inspecting, and operating Pixeltable applications.

App for Hyland Content Lake

Next-gen spatial workspace & mission control for autonomous AI agents (Hermes, Claude Code, Codex, Gemini). 3D living office, multi-agent DAG orchestration, zero-config hooks, and Obsidian knowledge vault.

基于 Yunzai 的 AI Agent 插件框架：LLM 对话·工具调用·长期记忆·MCP·工具进化 | A modular AI Agent runtime for Yunzai with self-evolving tools

Krineon: multi-agent, evidence-driven vulnerability-research harness

🏛 Multimodal 🛖 AI 🏘 Disaster 🏢 Prediction 🏣 is 🏥 an 🏦 advanced ⛪ research 🏩 project 🏪 exploring 🏭 the 🏯 use 🕌 of 🚞 Artificial 🚋 Intelligence 🚌 multimodal ✈ learning 🛸and 🚠 deep 🚟 learning 🚁 to 🛼 improve 🏎 the 🚂 prediction ⚽ early ⚾ detection 🥎 of 🏀 natural 🏐 disasters 🎮

Three AI research teams debate one stock into swing and long-term calls, on your own Claude or ChatGPT plan. Research only, not financial advice.

Multi-agent RAG with a critic agent that scores groundedness and flags hallucinations in real time. FAISS + sentence-transformers + Gradio.

Open reimplementation of STAIR: structure-aware generative retrieval that returns real section titles instead of chunk IDs. Constrained decoding gives 0.00% hallucination. Runs Mistral-7B QLoRA in 5GB.

Sleep-cycle memory management for coding agents — consolidate offline, verify against source, forget on schedule

Open-source AI agent for Ray-Ban Meta smart glasses. Runs your TV and laptop, sees what you see, remembers you. Gemini, Claude, ChatGPT, DeepSeek. Free.

DeepSeek Harness 的 QQ 渠道插件（OneBot v11 / SnowLuma）：工具面、长期记忆、知识库、主动行为与管理后台

EXSALA HERMES MOD - streamlined execution profile + skill pack for Hermes Agent

审计 agent 应用的 agent:每条发现附可重放 PoC,审计过程自身进哈希链。治理级 Runtime(stdlib-only)+ AGI 漏洞分类法 + 规则即数据引擎。An agent that audits agent applications.

Verification-first Claude Code harness by a mechanical design engineer — 13 validation gates, trust grades, heterogeneous-model cross-verification, supply-chain INTAKE, claim ledger

SWE-Ladder: Learning to Localize, Reproduce, and Patch via Progressive Privileged Self-Distillation.

Codebase Knowledge Graph & Architecture Intelligence platform. Transforms complex repos into interactive visual graphs with context-aware AI chat. Features fast AST parsing, hybrid BM25 retrieval, and multi-source Git/SSH sync. Self-hostable, runs locally, with all credentials encrypted at rest via AES-256 for complete privacy and security.

Desktop app to run Claude Code, Codex, Gemini CLI and OpenCode side by side, with a shared task board and memory over MCP, an orchestrator that runs workers in git worktrees, and Hangar Mobile to follow and steer them from your phone. Local, open source.

A runtime where LLM inference is the primitive unit of computation — conversations are processes, prompts are in-memory data, a system image is one JSON ｜ 以 LLM 推理为原始计算单元的运行时

A local-first multi-agent coding harness that connects model providers and ACP agents, binds them to specialized roles, and coordinates them in a shared session.

VergeCare: AI Property Caretaker helps you manage the home, film & spot repair issues, get a costed snag list, find handymen within your budget, track & automates bills payment. Agentic AI Caretaker for Property Managers, Landlords and Residents. Real Data from homes across the UK and US.

MeAI Vault - Permanent Agent Memory for Hermes

Nexa Agora — the workspace where people and AI agents run the company together. Connect the AI you already use through MCP; Nexa Bridge brings your own models and, next, Physical AI.

A shared knowledge and context workspace for humans and AI agents, with native MCP and REST support.

Agent Reliability Toolkit for LLMs - Test, evaluate, stress-test, and red-team your AI applications with scenario-based testing, multiple evaluators, and multi-provider support.

Reverse engineering with AI agents on Windows: verifying, measuring and visualising the tools agents use, with rea and Ghidra