Havuz

Models Take Notes at Prefill: KV Cache Can Be Editable and Composable (arXiv:2606.17107) — paper, code, results, and interactive companion.

Let Claude run code on your real machine — safely — from any Claude chat. Bridges Claude Cowork to Claude Code on your Mac/Linux box. One co

An end-to-end framework for Agentic Reinforcement Learning.

Autonomous pipeline that finds energy and CO₂ hotspots in any codebase and ships efficiency fixes proven by embedded statistical A/B testing

Put an AI agent on rails: mount a Burr state machine as an MCP server so the agent can only take the next allowed step, with every step reco

A high-performance, private, on-device AI chat client built with Flutter. AstraLM runs Large Language Models (LLMs) and Vision models 100% l

Open-source AI interior design agent. Turn room photos into scaled floor plans, wall elevations, photoreal renders and a design report where

Readable agent skills with optional declarative YON protocols you can inspect and validate. A personal, field-used pack by Alexandru Mares.

TypeSafe (Jev) skill routing for Hermes Agent: names the one skill worth loading, before the model call. Opt-in, stdlib only, ~$0.001 per ro

Open-source platform for internal AI agents. Your team's knowledge, rules and tools once, used in chat, workflows and your AI assistants.

Turn your repos into a team of Claude Code agents with persistent memory, coordinated tasks, and a live dashboard.

Run a flock of AI coding agents, each in its own git worktree, started by your ticket queue, merged by you. Work assigned to you in Jira, Li

Chromium build for AI agents and scrapers: 33 agent-level CDP commands, stable fingerprint identities, free Windows binary.

Error gates for AI coding agents: detects recurring tool-call failures and promotes them into enforced gates - remind first, hard-block on s

SciPIvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents

Counterfactual replay and causal attribution over LLM-agent trajectories: find which step actually caused a failure by intervening on it.

🐱 MiNerva·咪涅瓦 — 替你干活的赛博猫｜An autonomous AI agent that actually does the work.

Supervisor/Lead/Peer governance for multi-agent coding — one policy core enforced identically on Pi and Claude Code, brief-scoped authority,

Agent memory for Claude Code, Cursor, Codex: MCP server + Graph RAG over your Obsidian or markdown notes, a search layer for a Karpathy-styl

Zero-token long-term memory for Claude Code. Indexes your own session transcripts and retrieves evidence deterministically — entity-context

ChatGPT <-> Codex bridge: local coding agents ask ChatGPT for advice, and ChatGPT can inspect or steer your running Codex sessions. No API k

An MCP server built on top of Huawei Training Camp (HTC) APIs, enabling AI agents to access and analyze Huawei Health training data.

How to design AI agents that work as one: nine parts of an agent and its team, checked against 85 sources. Free guide in Markdown and PDF, w

Curated catalog of open, specialized AI agents for real business work — organized as broad umbrella categories (video, image, voice, SEO, so

MIT-licensed agent framework for the AxisVela ecosystem: governed agent loop, model/tool/memory/event primitives, plugins, sandboxing, and p

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

Installable graph engineering for Claude Code, Codex, OpenCode, Cursor, and OpenClaw — dependency-graph execution with local caching, qualit

A long-horizon CRPG benchmark for frontier agents. Raw 320x200 frames, Traditional Chinese, isometric navigation, twelve books to find.

DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks. Code, paper and blog post.

Spectrum: a deterministic semantic encoding enabling constant-time retrieval primitives

Stateful execution layer for browser agents.

Repairs predictable tool-call mistakes that often derail open weight models. Supports integration with other extensions' custom tools.

Multi-model consensus debate via the filesystem. LLMs propose, peer-review, rebut, vote and synthesize a group-confirmed answer. CLI + MCP.

Deterministic code-graph (GraphRAG) over your repo for LLM agents — local-first, git-native, zero-infra, served via MCP. Python, TS/JS, Rust

Single-agent multi-modal system for insurance damage claim verification. #1 globally, HackerRank Orchestrate June 2026 (15,295 registrants).

AI-powered Lead Qualification & Routing System built with n8n, Gemini and Python.

The open testing standard for voice AI agents. Deterministic + semantic + RAG augmented evaluation. Local first. Zero telemetry.

An unattended maintainer for one repository. It reviews, files, prunes, and merges only against proof that a guard bites.

Claude Code plugin that makes coding agents measurably cheaper over time: collect token costs, distill candidate rules, benchmark them on a

Production-ready Claude Code decision intelligence marketplace with adversarial routing, evidence-driven analysis, specialized agents, bench

CatchBench: a benchmark for auditing agent failures across the full lifecycle.

spec-builder + harness-bootstrap. Specs both you and the AI understand. An agent team tailored to your repo, derived from your contract and

Catches multilingual voice agents placing the wrong order. τ-bench style benchmark + CI gate for voice ordering agents in English and Hingli

一个将现实议题、规则机制与角色关系转化为可交互多智能体仿真/演绎世界的生成与运行框架。