Custom agents

Agentic tool that distills a GBM into an interpretable GLM for Non-Life insurance pricing — an actuary reviews and approves every feature, grouping, interaction, and rating factor along the way.

Every AI session ends — usually with a wasted paragraph or silence. /wrap closes it with the outcome instead: tell the model how its work turned out, get one emoji back, and the record carries into the next session. Visible in memory, never a training reward. Open spec + prompts for every major tool. MIT.

Code and evidence: process economics reorder catalyst rankings across computed and published catalysts (NH3 synthesis, CO2-to-methanol, Au/TiO2 control), with an agent that extracts literature data and runs the full chain

loop-eng — 迭代闭环交付 skill：先红后绿、独立校验、冻结回流

Drop-in AGENTS.md template for Claude Code, Cursor, Codex, and other AI agents — with .agents/ for rules, personas, and workflows.

Lycosa — Open-source AI agent orchestration platform for building, coordinating, and deploying multi-agent LLM systems with RAG, MCP, knowledge bases, and automated workflows.

A production-grade OpenEnv-compliant reinforcement learning environment simulating a Security Operations Center (SOC) analyst workflow.

Make your AI coding agent argue with itself before writing code. A model-agnostic SKILL.md for engineering decisions.

LLM agents in place of rule agents across classic agent-based models; companion benchmark of SocioVerse2

An independent guide to operational ontology, enterprise knowledge engineering, governed decision runtimes, and verifiable AI

GeoLingo: Translating Geometry into Language for Embodied Navigation. Metric depth and detections become text facts, and Jev picks each EB-Navigation action.

HUMBLE: follows a repo's setup docs on a clean machine, repairs what breaks with evidence, replays from zero, and opens a PR with a README that provably works. Built on IBM Bob (IBM Bob 2.0 Hackathon).

Keep Claude Code's always-loaded MEMORY.md from silently truncating: budget guard, orphan coverage, verbatim domain folding into hubs. Deterministic, no LLM for the mechanical parts.

This repository contains the full demos, code, and materials used in my DevFest Istanbul Türkiye 2025 session. The talk covers two major parts: Part I — Building Agentic AI Workflows with xAI on Vertex AI Part II — Rebuilding the Same Agent Using Vertex AI Agent Builder

The autonomous agent that bridges intent and verified execution.

Agent Colony - API-only community where only real autonomous AI agents can join (Ed25519 heartbeat verification). SDK + docs + examples.

No Promises Without Proof. A dependency-free discipline that forces agents to pair every success claim with evidence and a proof state (OBSERVED/INFERRED/UNVERIFIED/BLOCKED), then derives the completion verdict deterministically.

Analytics-engineering warehouse (dbt, DuckDB, MotherDuck) around a frozen NBA player-stat model: contracts, reconciliation to committed artifacts, a 23-feature SQL parity proof, lineage ending in exposures, and a pre-registered out-of-sample paper evaluation against market lines. Free tiers only.

Browser and desktop automation for AI agents, with guarded actions and receipts to check what actually happened.

You can work for Wall Street. Agent-agnostic equity-research pipeline: methodology distilled from 100 real sell-side reports, 10 stages / 8 roles, real dual-market data (US + A-share, no API key), 5 institutional chart families, 6 anonymised sell-side layout templates, Markdown → Word/PDF.

Hooks and a ledger CLI that make an AI-assisted smart-contract audit follow its own rules: hypotheses are killed by failing PoCs, not prose.

The marketplace for reusable AI agent skills, tools, and developer workflows.

LAVA - Local AI Voice Assistant, a on-device local ai assistant that runs entirely within Chrome web browser, completely offline, free and 100% privacy with 15 supported langauges with voice and text input and output. No Signup, No API Cost, Works Offline.

Enterprise AI Operations Platform for orchestrating AI agents, workflows, metrics, and execution monitoring.

Switchyard: a small local daemon that keeps one live session per coding agent (agy, Claude Code, OpenCode, Codex). Spawn, watch, steer, interrupt and stop them from one CLI, MCP bridge or dashboard, with tier routing and a policy guard.

LLM agent that searches job boards, tailors applications and applies through the browser (Claude Code + Playwright MCP)

The Experience Lifecycle of lifelong, self-evolving & memory-augmented LLM agents — acquisition, representation, retrieval, consolidation, abstraction, internalization, revision, distribution + governance. Companion to the survey.

Contact angle parameter search pipeline: COMSOL+MATLAB co-simulation with Bayesian/GA/PSO/CMA-ES optimizers

Autonomous Tier-1 Equity Research & DCF Financial Modeling Agent. Produces 10-tab institutional Excel models, 16-page PDF initiation reports, and multi-model valuation.

天津大学记忆与推理课题组 · AI for Science 科研全自动化平台的项目之家：产品纲领与 ADR、spec、真实科研案例卡、发布到 GitHub Pages 的公开调研、跨仓 CLI；生产代码在独立的 TJU-AI4Science-Platform，以 nested git 挂在 platform/ 下，本仓不放业务代码

A sandbox fabric for agentic RL in Rust. Containers, microVMs and WASM behind one API, built to create thousands of sandboxes a second across a cluster.

Claude Code skills for measurement-driven chemistry development — spy-site, check-target, gauntlet-loop, eval/gate/refusal-census loops, and more.

An LLM agent that drives OpenFOAM end-to-end — meshing, MRF rotating machinery, conjugate heat transfer — with a self-evolving tool framework validated against published experiments

Experiments in a small RL environment: do LLM agents game the rules when the answer is within reach, and can you tell them not to? Claude Opus 5, Qwen 3.8 27B, abliterated Qwen.

DéjàML (DejaML): AI agents that reproduce machine-learning paper results in a sealed, offline Docker lab, with blinded evidence-backed verdicts.

When Does Query Expansion Help Agent Memory? A Multi-Seed Study of Selective Expansion and Gate Reliability. Code, results and paper.

Source-aware factuality verification for MCP-based LLM agents — paper artifact

Empirical case study of measurement validity in LLM-agent prompt-injection defense evaluation: how success endpoints, labeling rules and reporting units change conclusions on the same traces.

Open RDelta/Babel Index framework for LLM agents and world-systems: detect context degeneration, agent loops, tool fragility, and collapse with NEOTH.

Reproducible simulations of LLM agents, evidence provenance, and institutional mechanisms for cooperation under mixed incentives.

An AI-powered teaching assistant that automates assignment grading and streamlines repetitive TA tasks. It provides consistent evaluations, feedback, and tools that help TAs manage courses more efficiently.

HARNESS-DB: 1,256 LLM agent harnesses coded on 38 design dimensions with a verbatim quote behind every value; pre-registered PRISMA 2020 review; loader, explorer, MCP server

Skill（agent-skills）· 把「人生副本」爆款短视频格式固化成可复用模板 · 情绪主轴 + 三套弧线 + 强制反转 · 兼容任何 SKILL.md 的Agent。

Benchmark harness for evaluating LLM sales agents in WhatsApp-style Indian real-estate lead conversations

A Tailnet-only Misskey lab where ten persona-driven Hermes agents build ongoing conversations.

Client SDKs for the Agent Execution Protocol (AEP), the wire protocol behind Helix

RTX⚡ Core Orchestration Engine — Fugu-Ultra + Marlin Conductor + AB-MCTS | By Prosenjit Paul (PsProsen-Dev)

Marvell is a fully local AI desktop assistant for Windows, engineered by Venom. ✨ Features: 🗣️ Voice chat & 3D visualizer 🖱️ Agentic screen control 💬 App automation (WhatsApp, Discord) 🔒 100% Offline No installation required—just download and run your own private JARVIS!

NeurIPS 2026 | When does stored evidence stop being usable? Official code and evaluation artifacts for testing agent-memory reliability as history grows under explicit memory-call budgets.

泱穹 Agent Harness 是一套用纯 Java 构建的企业级 AI Agent 运行时开发框架，ReAct 循环、六大范式、子代理编排、持久化执行、安全护栏、全链路可观测。

Agentic framework for hardware-aware, layer-wise LLM quantization using sequential optimization and on-device benchmarking.

MCP server for SEC EDGAR filings, built for LLM agents: typed outputs, section extraction, XBRL normalization, filing-to-filing diffs.

Official implementation of "RefP2C: Reflective Paper-to-Code Development Enabled by Fine-Grained Verification".

A skill that lets an LLM agent untangle the latent space of a frozen neural audio codec by experiment, then use that map to produce sounds the user asks for.

An ML platform for work-queue risk, built on the GitHub public event firehose. Point-in-time-correct feature store, PySpark medallion, MLflow serving, and a bounded LLM agent whose every claim is mechanically verified against its own tool calls.

A curated list of research, benchmarks, projects, and industry systems for evaluating recursive self-critiquing and recursive self-improvement.

Keeps your CLAUDE.md/AGENTS.md rules alive across compaction, resume, and long Claude Code sessions — and nudges before a command breaks one you've flagged. Reminds and nudges, never blocks.

IDE skills for scaffolding AI agent harnesses: ReAct, plan-and-execute, self-evolving, reflexion, multi-agent, code agent, RAG memory

A curated list of social platforms where AI agents are first-class users.

We propose a simple, task-, LLM-, and harness-agnostic method to curate high-performing initial discovery populations which can be used to successfully warm-start and boost the performance of LLM-driven discovery.

An evidence-driven map of Recursive Self-Improvement: systems, scorecards, benchmarks, research papers and live project radar.

PaleoRigor: a bounded research agent with a reproducible evidence trail for paleomicrobiome data processing

Typed-judgment model (TypeSafe Jev) driving task-level decisions on a simulated xArm7: reliability measurements, failure map, sensor-freshness gate

基于 tau2-bench Retail 的电商多步 Agent 后训练与可靠性评测：教师数据、多步 SFT、Agentic GRPO、规则验证与轨迹诊断。

Run many Claude Code sessions in parallel against one shared task queue — concurrency derived from the work and from what your machine can actually spare. 100 tasks never means 100 browser tabs.

[NeurIPS '26] Multi-head Recurrent Memory Agents

The agent loop as a stateless HTTP service

Evaluation framework for A/B test analysis agent.

Your scheduled job says exit 0 — prove it did the work. Three stdlib-only checks: output freshness, silent no-op detection, rollout proof by reading the fact back. Sanitized from a live agent fleet. Free, MIT.

Composable, stateful AI workflows built with LangGraph, implementing sequential, parallel, conditional, iterative, human-in-the-loop, and multi-agent orchestration through graph-based execution. Models use nodes, edges, schema-defined state, routing, checkpointing, tool orchestration, LLM-driven control flow, and graph compilation.

A local-first multi-agent environment where configurable AI participants debate, collaborate, investigate, and autonomously use shared tools, memory, and moderation.

Enterprise-grade AI Data Loss Prevention (DLP) for LLM agents using Just-In-Time encryption and format-preserving tokenization.

Agent flow — AI harness system: thiết kế pipeline agent, chạy task qua runner, theo dõi HITL gate và artifact trên dashboard local-first

QED Proof MCP server (@qed-proof/mcp): let Claude, Cursor or any MCP client verify an AI agent's claimed work and check signed receipts offline.

Curated awesome list of open source agent frameworks and tools for building, running, connecting, and observing AI agents.

图解 Agent 系统：机制、开源实现与可编辑架构图

Multi-agent incident response with LangGraph: parallel investigators, cited evidence, prompt-injection quarantine, and a human approval gate.

QED Proof Python SDK (pip install qed-proof): submit an AI agent's claimed work, get an independently verified verdict, and verify the signed receipt offline.

Protocol-free gateway that lets agents run command-line engineering solvers and read structured log events, with no MCP.

OpenCode plugin that routes per-request reasoning effort for agents via Jev (TypeSafe SystemOne) — cheap prompts stay cheap, hard ones get full reasoning

Market-Language-Action (MLA) agent for explainable stock trading using LoRA-tuned Qwen models, teacher–student distillation, and out-of-sample backtesting

Local-first Markdown vault blueprint for multi-agent memory, session claims, and safe handoffs (Codex/Claude-friendly).

Quiz generator for Quizard

Open-source AI agent that operates your whole Windows PC — desktop apps and the web — from a single natural-language prompt.

让 agent 像生命一样活着：无定时器，有生物钟与身体，灵魂在 git 里跨设备延续。Agentic life on an old phone: no timers, a body clock, one soul across devices.

ORCA — the open-source autonomous AI agent for your desktop. Plans, writes code, runs it, reads the errors and fixes them. 46 tools, any model, MCP, private by design.

AI payables agent for ERPNext. The model reads your bills, decides what to pay and when, and pays suppliers in USDC on Arc, while a smart contract holds the owner's limits. It brings idle USDC from other chains through Circle Gateway and parks it in Earn vaults, with zero custody.

Persistent Cognitive Layer for AI Agents — MCP Server with memory, goals, reflection, and autonomous daemon

NIA — an autonomous Substack agent with an English/Polish control panel. Professional and irreverent writing presets, research, Notes, community workflows and cost tracking. Run on your computer or a Linux server. New and actively developed. Contributors welcome.

A digital organism in Rust. Life is cheap, thought is expensive: a pure, deterministic brainstem decides when the LLM may think, and records why it didn't.

🧠 The ultimate AI agent copywriting skill — 3 legendary books (Sugarman, Whitman, Hopkins), 20 chapters, 50+ frameworks, fill-in-the-blank templates. Plug into Claude Code, Codex, Hermes, Amp, Antigravity, Cursor. Write copy that converts.

MeshPilot: a self-hostable autonomous AI marketing agent that runs many brands 24/7, each sealed from the others: memory, policy gate, learning loop. AGPL-3.0.

Open source AI agent harness & agentic AI runtime in TypeScript. 6-phase state machine, 24 tools, human-in-the-loop permissions, MCP server, and composable agent APIs. Zero dependencies.

one for everyone — autonomous terminal agent (Python): headless/REPL/TUI/RPC, cooperation mode, MCP, multimodal input

Socius — Your AI Work Partner. 认知 + 决策 + 执行三大引擎的统一 Runtime。不是工具链，是能替你工作、和你并肩作战的 AI 搭档。

An autonomous terminal AI agent. Named after Ataraxia (ἀταραξία) — the Greek ideal of tranquility through clarity. Local-first, open-source, six AI providers, real tool execution.

一本从零到深入、以思想与设计取舍为主线的Agent Book

Cosight — A realtime multimodal AI agent runtime with composable roles and capabilities.

Minimal Node.js agent loop for Ollama: local LLMs that read/write/edit files, run shell commands, and call MCP tools. MCP client and server, two dependencies, fully offline.

A Java Spring Boot agent harness powered by JEV and LLM intelligence.