31 October 2026

Persistent memory across sessions for coding agents

Captures what an agent does during sessions, compresses it with AI, and injects relevant context back into future sessions.

• Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, and OpenCode.
• The README calls it a persistent memory compression system built for Claude Code.

Open-source Grok Bot alternative with a virtual machine

A team of AI bots inside a chat app. Each bot is a local Claude or Codex agent with its own model, cloud computer, and connected apps.

• An independent project inspired by Grok Bot, not affiliated with xAI, Meta, OpenAI, or Manus.
• Bring-your-own-agent and local-first, on the models you already have.

The design vocabulary for your coding agent

Commands like polish and bolder steer design, critique and audit review it, and AI slop gets caught before it ships.

• Install with npx impeccable install from the project root, then run /impeccable init in your agent.
• Works with Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, OpenCode, and more.

Open-source AI coworkers, each with a computer of their own

An AI assistant on your infrastructure. Each coworker gets its own browser, files, and tools; every action is decided before it happens and recorded after.

• Supports any agent stack through AG-UI.
• Only the tools you grant are available, and everything can be changed because it runs on your own infrastructure.

Open-source chart engine for Zi Wei Dou Shu

A chart engine for Zi Wei Dou Shu, a traditional Chinese destiny system, based on Ni Haixia's teaching, with a full algorithm and a pattern knowledge base.

• The v4.0 interpretation dataset holds 674,976 charts across 13 dimensions, 8,774,688 segments in total.
• The unpacked dataset is 56.6 GB, released on Oct 15, 2026.
• The README is in Chinese, with an English version.

Official AI agent skills from the Expo team

Skills that teach AI agents Expo and EAS so they can build, deploy, upgrade, and debug Expo apps.

• Expo documentation, Expo CLI, and EAS CLI remain the source of truth.
• Claude Code or Codex install it as a plugin; Cursor and other agents use the skills CLI.

A unified team workspace with shared AI memory

Brings email, chat, docs, tasks, agents, calls, and CRM into one interface, @-linked together.

• Everything in the workspace is linked and searchable by the team and its agents.
• Built to replace a mix of Slack, Linear, Notion, HubSpot, and Superhuman.

Skills for real engineers

Agent skills the author uses every day, straight from his .agents directory.

• Designed to be small, easy to adapt, composable, and to work with any model.
• The author contrasts them with GSD, BMAD, and Spec-Kit, which own the process.

Control your coding agents from phone, web, and desktop

An agent harness control surface for the agents on your machine, through iOS and Android apps, a web app, and an Electron desktop app.

• Uses your existing Claude Code, Codex, Cursor, Grok Build, OpenCode, and Google Antigravity subscriptions.
• The makers say they sell nothing and wanted a fast, remote-ready, truly open experience.

Lifelong personalized tutoring

A personalized tutoring system for long-term learning. Release v1.7.0 (Oct 28) adds modular installation, a refreshed learning workspace, and narrated video lessons.

• Block notebooks with Dito and resumable embedding checkpoints arrived in v1.7.0.
• Earlier releases added French, Ukrainian, and German interfaces and question-bank practice.

Anthropic's public repository for Agent Skills

Skills are folders of instructions, scripts, and resources that Claude loads dynamically for specialized tasks.

• Examples include creating documents with brand guidelines or analyzing data with an organization's workflows.
• Holds Anthropic's skills implementation for Claude; the Agent Skills standard lives at agentskills.io.

An agentic IDE that fits the way you work

A lightweight IDE with an interface around the pi coding agent, using isolated workspaces and living project specs.

• Each workspace is a git worktree with its own branch and directory, so agents work in parallel without touching your checkout.
• Full extension support is coming so ThinkRail can create the tools it needs.

A collection of system prompts and personas for many models

A collection of 32 prompts for 26+ models: system prompts, persona definitions, workspace policies, AGENTS.md files, and Ollama Modelfiles.

• Covers open-weight families such as GLM, DeepSeek, Kimi, Qwen, Gemma, and GPT-OSS.
• Also covers closed frontier models such as Claude, Gemini, Grok, and Mistral.

A software development methodology for coding agents

A complete methodology built on composable skills, plus initial instructions that make sure the agent uses them.

• The README has install guides for Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, and more.
• It includes a Basic Workflow section and a guide for when something goes wrong.

The headless browser designed for AI and automation, written in Zig

A new browser built from scratch for AI agents and automation, not a Chromium fork or WebKit patch.

• The README claims 16x lighter and 9x faster than Chromium.
• Benchmark of 933 real web pages on an AWS EC2 m5.large: 123MB peak memory vs 2GB, 100 pages in 5s vs 46s.
• Install with Homebrew, AUR, nightly builds, or Nix.