Code agents

Trusted architecture context for AI agents: an evidence-backed architecture graph from OpenAPI, AsyncAPI and OpenTelemetry, served over MCP.

MCP server for the GoModel admin API — grouped tools with gradual discovery, live docs from GitHub, TTL read cache, opt-in HTTP host mode. Runs on Bun; distroless Docker; auto-releases via CI through gated PRs.

Persistent engineering memory & session-continuity control layer for AI coding agents (Claude Code, Codex, Cursor, Gemini), built on the Model Context Protocol. Scans a repo once, remembers architecture/decisions/changes, checkpoints safely, and hands the next agent an exact continuation brief — so agents stop rescanning and losing context.

面向 A 股的日频量化研究与回测引擎，支持传统因子 + ML推理 + Python函数混合策略

Scrapling 风格隐身爬虫库：自适应选择器 · TLS 指纹 HTTP · JS 渲染 · 回调式 Spider 框架（重试/robots/去重/中间件管道）· JS 逆向 Agent · 验证码识别 · MCP 服务 · 本地 Web UI · 授权渗透侦察 · 默认 SSRF 防护

A security-conscious, local-first LLM agent in Go powered by Ollama & Model Context Protocol (MCP). Features defense-in-depth sandboxing, tamper-evident audit logging, and MCP tool orchestration.

梁神按需模式 / LiangShen Lazy — DeepSeek Harness 的按需工具隔离 preset：默认极简，需要什么能力才临时开放什么，用完自动收回。A minimal-by-default, capability-on-demand agent preset for DeepSeek Harness.

A multi-agent research system for the humanities and social sciences.

T2AG 是考你而不是替你答的 AI 学习框架：错题、证明与每一次裁决都存在归你所有的纯文本里，任何 agent 可读可带走；环上的裁决者是你。An AI learning harness that tests you instead of answering — your mistakes, proofs and verdicts live in plain text you own, portable across agents; the judge in the loop is you.

Open Source Desktop AI Agent Runtime - LLM, MCP, Voice Runtime, and local/cloud models wired into one PyQt6 desktop agent (Ollama + Hermes)

A progressive source-study tutorial of DeepSeek Harness (dsh): 17 chapters from a minimal agent-loop to multi-agent delegation, compaction, permissions and cross-process RPC, with runnable teaching code. Zh/En.

A full agentic loop delivered to your phone

Context-driven valuation bias and halo effects across six multimodal LLMs (companion study to Lee, 2026)

Six patterns for working with AI agents: multi-agent, agentic workflows, Claude Code. Traps and cost of each. ES + EN · Seis patrones para trabajar con agentes de IA

UI and workbench for AI trading agents: run TradingAgents (LLM multi-agent trading framework) on a real stock, watch every agent live, keep a database of runs, benchmark LLMs, backtest honestly, and send the decision to an Interactive Brokers paper account.

A tiny, stateless API that gives OpenAI, Claude, Gemini, xAI and Ollama one simple text-in/text-out contract.

JMP — Joint Model Participation. A local coding workspace where Jev routes actions and OpenAI, DeepSeek, or local models generate arguments.

Post-response grounding check + cross-family adversarial review for LLM chat agents

LLM-driven browser agent: observe → plan → act loop on Playwright + FastAPI, live-streamed to a React UI. No agent frameworks.

A deterministic context-driven agentic SDLC framework for Cursor and Claude. 41 codified laws and 22 specialist roles to eliminate LLM hallucinations and context drift.

A cockpit for the agents that write your code.

Hermetic sandbox executor that returns typed, delta-aware feedback to coding agents (Rust). Parsed compiler/sanitizer/test verdicts + run-to-run deltas instead of raw stdout.

Recursive Language Models for pi: persistent Python REPL with rlm() sub-agent recursion, namespace snapshots and a continual harness (/refine) — as a single extension

Take the pain out of installing and running machine-learning interatomic potentials: every major MLIP in one registry, set up, verified and run by your LLM.

An autonomous multi-agent orchestration harness. Agents coordinate through one SQLite-WAL authority — claim work, hand off bounded context, assign audits, run tracked local model jobs, and resume across multi-day horizons — with federated retrieval over large multi-domain repos and read-only web, macOS and iOS clients.

Autonomous, read-only threat-hunting platform for SOC analysts: an LLM agent turns a hypothesis into sourced IOCs, hunts across Microsoft Sentinel, Defender and Google SecOps under human checkpoints, and delivers an investigation report.

A curated list of code world models: executable dynamics, programmatic simulation, and coding agents that build and reason about worlds.

Open source runtime to migrate DRF to FastAPI (and more to come) 🚚

Safe Multi-Agent Deployment Platform: pairwise safety analysis for autonomous agents that share a filesystem, clipboard, and OAuth scopes. Every verdict is graded on a four-rung evidence ladder, from docs-only to executed-in-sandbox.

CogniTrace bridges the gap between structured clinical tracking and compassionate, daily memory care. Built on FastAPI, Next.js, and Google Gemini, it provides a dual-interface ecosystem tailored specifically to the unique needs of Caregivers and Individuals experiencing cognitive decline.

A Retrieval-Augmented Generation chatbot that impersonates Claude Shannon, the father of Information Theory. It ingests his biographies and papers, retrieves with hybrid BM25 + FAISS, reranks via cross-encoder, and answers in his voice (English/Persian).

Full-stack Agentic AI platform demonstrating seamless database-to-LLM integration. Built with Gilhari ORM, ORMCP, FastAPI, LangGraph, and a Flutter Web dashboard featuring 11 Generative UI widgets.

🪙 40+ Free LLM APIs → One OpenAI-compatible endpoint. CLI, TUI, Proxy, LangChain wrapper. Auto-rotate keys, handle rate limits, track usage.

SQL intelligence layer for LLM agents — phase-locked generation, property graph schema, 10 dialects, MCP server

smart-contract-reviewer is an open-source LegalTech tool that automatically checks the compliance of permanent employment contracts with Moroccan Labor Law. Leveraging Generative AI and RAG frameworks like Gemini and LangChain, it detects potential legal issues, suggests corrections, and ensures that contract clauses adhere to local regulations.

神光 Agent 课程的陪练式学习仓库。内置 course-study-coach skill,每课自动走建目录→陪练→提问查漏→生成笔记 6 步,基于学习科学(ZPD/主动回忆/费曼学习法)。欢迎 fork ｜star ✨

Local-first AI reasoning observability for agents: record runs, diff golden baselines, and fail CI on path drift. Complements OpenTelemetry and LangSmith. Python port of DProvenanceKit.

Python SDK for HelixID — onboard agents, sign and verify verifiable presentations through the API, and delegate authority, with LangChain, CrewAI and MCP middleware integrations. Mirrors the JS SDK's surface.

Deterministic AST linter for multi-agent systems. Finds unattenuated agent delegation. No dependencies, no network, never executes the code it reads.

Browse a curated list of AI slop projects and see examples of broken, risky, and absurd AI-generated code

Production-grade multi-agent orchestration on LangGraph — Postgres-backed durable state, dual-mode JWT + Azure AD SSO, per-team isolation, and a team-wide review queue. Advanced (LangGraph) tier of the Multi-Agent Orchestrator Template family.

Claude Code plugin with 27 specialist agents, web verification, Red Team review and a final quality gate that blocks weak or unverified answers.

AI 业务分析与操作手册生成专家 | Intelligent business analysis & user manual generator. 六层递进 + 知识图谱 + 增量回灌

Durable workflows for agent teams: named roles, reviews that send work back, a person at the merge. Build a software factory or a domain-specific harness on the engines you already use. No server, no database.

Ninebrains: one Brain, many thinking arms. Open-source desktop workbench for parallel Claude Code and Codex agents with verification gates.

Parallel OpenAI Codex agents inside Claude Code — multi-agent orchestration with git worktree isolation, resumable sessions, and dual-model PR review. MCP plugin.

An AI-powered autonomous data analyst that accepts CSV input, generates Python code using LLMs, executes it, and returns computed answers with built-in retry logic.

A three-tier training curriculum and ready-to-copy pipeline for building multi-agent AI workflows in Cursor and Claude Code.

A file-based kanban board that AI coding agents and humans work together, with isolated workspaces and a parallel, model-agnostic dispatch workflow

Reliability Harness v2 for OpenCode CLI — task contracts, least-privilege sub-agents, clean-checkout verification, and a deterministic done-gate. Orthogonal agents (scout/spec-critic/test-designer/lead/verifier/adversary) instead of 3 identical thinking clones. Compatible with Oh My OpenAgent. Not a model swap or 1:1 emulation.

Local-first multi-agent orchestration — any agent on your machine can be the brain and dispatch to the others (Codex / Claude Code / Antigravity).

One model writes the work, a different one reviews it, and neither decides whether it passed. Cross-provider AI review with verifiable evidence, for DeepSeek Harness/Cordis hosts.

Cross-harness SupervisorLLM plugin for Claude Code, Codex, Cursor and Gemini CLI with native agents, hooks, web verification and final release gates.

Provider-neutral Rust coding agent with user-owned provider/model routing, tools, swarm, MCP, sessions, and cost tracking. Public beta POC. Proprietary.

多 Agent 协作打 CTF 实战手册：协作体系 / 纪律卡 / 技术套路库 / 已判死清单 / 实战复盘

A file-based dispatch protocol for multi-CLI AI coding workflows. Prompts are advisory; validators are authoritative. Uses markdown files for managing state, backlog, and handoffs.

Antigravity-style multi-agent orchestration plugin for OpenCode. 10 agents, 6 patterns, DAG engine with worktree isolation, typed artifact bus, cost tracking.

Spec-driven multi-agent Codex skill pack with /agent-do commands for planning, execution, reviews, security, design gates, and cross-platform coordination.

🤖 AgentChat 9AI — 9个免费网页AI的Claude Code技能套件（Gemini/ChatGPT/Claude/千问/Kimi/MiniMax/MiMo/DeepSeek/豆包），串行降级+并行编排+深度管道

The complete collection of OpenAI Codex skills, automations, multi-agent workflows, and prompts. 50+ production-ready skills, 100+ reusable prompts, background automations, and battle-tested patterns for AI-powered development.

Deep Agents + LangGraph multi-agent research workbench: async sub-agents, hybrid RAG (Milvus), Mem0 memory, E2B sandboxes & Skills deliverables, HITL interrupts & cron — React + FastAPI full stack.

Parallel multi-agent skills for Claude Code — review, research, investigate, and implement with concurrent specialized agents

Claude Code skill: one session coordinates, spawns and reviews builder sessions across a multi-phase build

🦖 Four AI coding CLIs (Claude, Codex, Kimi, Gemini) debate a task, reach consensus, and combine to build it — using your subscriptions, not API keys.

Aria — an open-source AI operating system in your browser: a macOS-style web desktop with a live multi-agent brain (local LLM via WebGPU, real tools, live code execution).

Shared situational awareness for multiple coding agents working in the same repo

An autonomous Windows hardening agent with a local language model. It decides with an embedded model, fixes only what it can undo, verifies every change, and rolls back anything that made the machine worse.

Agent-driven code intelligence engine: Python/FastAPI agent orchestration, Rust/Axum code graph and context compression, plus a React + Vite WebUI. Early-stage, work in progress.

Contextra is a high performance AI context engineering platform built in Rust for document ingestion, retrieval, embeddings, memory, prompt orchestration, and multi-LLM application development.

Describe an AI agent, get a ready-to-run agent home with persona, memory and a growth loop. Build one agent or a whole collaborating team.

Agentic AI red-team - autonomous decide->execute->observe->reflect loops against own-lab targets, human-in-the-loop safety gates.

Lightweight, high-performance AI assistant for Telegram built in Go — self-hosted workspace operator with local tools, skills & smart memory.

A self-hosted AI assistant that can rewrite its own code — Claude Code CLI behind a chat UI, with crons, webhooks, connectors and generated apps.

A native desktop IDE for Lean 4 on macOS, Windows and Linux. Every proof re-checked by Tenet, an independent Lean 4 kernel. An MCP server for AI assistants, with Git and GitHub built in.

transformers gguf text-generation-inference llama-cpp omnimergekit merge

transformers safetensors qwen3_5 image-text-to-text text-generation-inference pytorch

One agent config, installed into every coding harness

Unofficial zero-dependency Taiga.io CLI and AI agent skill. Token-efficient TOON output for LLM context windows. Python stdlib only.

Claude Code skill — guided interview that recommends AI skills for your repo or team. Install with one prompt.

A secure, highly capable MCP server written in Rust. It exposes a comprehensive suite of filesystem operations as tools for AI agents.

transformers tensorboard safetensors phi4mm text-generation generated_from_trainer

The AI that remembers you. Forever. — A living digital entity built in Rust. 68 crates. Self-writing genome. Persistent memory. Constitutional governance. Drop a TOML file, Hydra learns.

Open-source Rust runtime for autonomous agents — 13 modules, 76 crates, 2625 tests. Event-sourced persistence (Lago), homeostatic regulation (Autonomic), x402 payments (Haima), cryptographic identity (Anima), distributed networking (Spaces), metacognitive evaluation (Nous), OpenTelemetry observability (Vigil).

Long-term memory for AI coding assistants. Rust CLI with semantic search, Agent Skill integration, and local deployment.

A deterministic pipeline that diagnoses, explains, and safely auto-fixes failures in LLM agent systems.

AIcompanion è un assistente AI modulare sviluppato in Python, progettato per elaborazione conversazionale, analisi documentale e modalità di interrogazione. Integra Flask per l’interfaccia web, LangChain per la gestione del contesto e modelli Ollama eseguiti in locale fornendo un ambiente completo e ottimizzato per workflow AI on-premise.

A sophisticated, AI-powered trading system that combines traditional technical analysis with modern machine learning, multi-agent AI systems, and intelligent risk management - all wrapped in a beautiful, mobile-friendly dashboard! 🚀

Multi-agent AI team builder for OpenClaw — bootstrap complete teams with auto-dispatcher, dashboard, and infrastructure

Fast, native Go CLI for Linear -- 35+ commands for issues, projects, cycles, initiatives, webhooks, and more. Built for agentic workflows and psst-compatible.

Local RepoGPT using Ollama + FAISS (RAG) + FastAPI Web UI. CPU-friendly AI assistant for querying code repositories.

Application for explaining difficult to understand textbooks based on Yandex Cloud API (YandexGPT-Pro, Yandex-OCR)

Minimal AI agent in your terminal

AI-first mail and workspace suite for Gmail/Outlook, calendar, contacts, tasks, and follow-up. Smart mail first, powered by local CLI agents.

Connect AI assistants to your YNAB budget for natural language expense tracking, budget analysis, and financial insights. Secure, fast, and easy to integrate.

Next.js app for Claude code best practices built with Fumadocs. Document and explore content sources, routes, and MDX docs for fast, scalable docs sites. 🐙

一个自动化密钥检测工具，对文件进行分析并提取潜在密钥。本项目获得DataCon2025软件供应链安全赛道第5名。

Multi-zone disaster relief AI env for Meta PyTorch OpenEnv Hackathon. 4-stage pipeline: PyTorch ZoneScorerNet -> Triage -> Planner -> Action Agent. False SOS detection, cascading failures, airlift precision.

Exploring the cognitive architecture and design of a multi-agent system leveraging Large Language Models (LLMs) for scalable workflows.