4 November 2026

Cross-session memory for AI coding agents

Records what an agent does during a session, compresses it with AI, and injects relevant context into later sessions. Works with Claude Code, Codex, Gemini, OpenCode, and more.

• Captures tool usage observations, generates semantic summaries, and makes them available to future sessions.
• The installer has an --ide option for OpenCode, T3 Code, Antigravity CLI, and OMP.

Computer use CLI for AI agents via the OS accessibility tree

A CLI and C library that reads and operates desktop apps through the OS accessibility tree, with no screenshots or pixel matching.

• A snapshot returns JSON with a ref for every element an agent can act on.
• There is no model inside; the read, click, and re-check loop runs in your agent.
• Available for macOS, Windows, and Linux.

CLI for Git worktrees, built for parallel AI agents

Worktrunk makes worktrees as easy as branches: they are addressed by branch name, with paths from a configurable template. Designed for running many AI agents at once.

• wt switch -c -x claude feat creates a new worktree and starts Claude in it.
• It has hooks to automate local workflows and copy-on-write build caches.

C inference engine for huge MoE models on ordinary PCs

colibri runs mixture-of-experts models with hundreds of billions of parameters by keeping hot experts on the GPU, warm ones in RAM, and streaming the rest from NVMe.

• Written in pure C, one file per model family; a GPU is optional (CUDA or Vulkan).
• Needs at least 8 GB of RAM and 22 GB of free disk for the smallest model.
• It also reads already published GGUF files.

Agent skills for real engineers, from the author's .agents directory

Agent skills the author uses every day for software engineering. They are small, easy to adapt, composable, and work with any model.

• Available as a plugin for Claude Code, Codex, GitHub Copilot, and Gemini CLI.
• Through skills.sh, editable files are copied into your project and updated by hand.

Training-free modification of language model behavior

Heretic changes the behavior of LLMs you can run yourself, with no training, using mathematical techniques based on contrasting example prompts and scorers. Its default action suppresses refusals.

• .toml configs exist to reduce slop, humor, or sycophancy, and to alter emotional alignment.
• It supports 4-bit quantization through bitsandbytes to save VRAM.

Spotify Portal plugins for Claude Code, Codex, Cursor, and pi

A plugin that brings Spotify Portal into coding agents through the Portal CLI, with workflows for authentication, catalog search, and invoking actions.

• Searches the software catalog and documentation with natural language, and builds service briefings.
• Portal actions are invoked with built-in help, dry-run, and confirmation safeguards.
• The marketplace also ships the shunt plugin, which routes I/O-heavy work to cheaper models.

Free local AWS emulator, a LocalStack alternative

An open-source local AWS emulator for development, testing, and CI with no cloud account or auth token. Point the AWS SDK, CLI, Terraform, or CDK at http://localhost:4566.

• Described as a drop-in LocalStack replacement: swap the image and keep your workflows.
• A standalone floci-aws binary ships for Linux, macOS, and Windows, with separate Azure, GCP, and OCI emulators.

Xianyu listing monitor built on Playwright and AI, with a web UI

A monitoring system for Xianyu (Goofish) listings that uses Playwright and AI models to analyze items, with a full web admin UI.

• Each task has its own keywords, price, filters, and AI prompt; tasks can run on a Cron schedule.
• Notifications go through ntfy.sh, WeCom, Bark, Telegram, or Webhook.
• Runs with Docker; the web UI is at http://127.0.0.1:8000.

Local voice dictation for Mac, Windows, and Linux

Hold the shortcut, speak, then let go: Hex transcribes on your device and pastes the text into whatever app you are in.

• The first run downloads a speech model; after that nothing leaves your machine.
• Experimental features live in Labs, at the bottom of the sidebar.

Web scraping framework from a single request to a full crawl

A Python library that combines fast HTTP requests, stealth browsers, parsing, and concurrent crawling. Use it from Python, the terminal, or an AI agent through its MCP server.

• Fetchers handle JavaScript and Cloudflare Turnstile; the parser supports CSS, XPath, and regex.
• Spiders have proxy rotation, retries, pause/resume, and JSON, JSONL, CSV, and XML export.

Open-source foundation model for financial candlestick data

Kronos is a family of decoder-only Transformer models pre-trained on K-line sequences from over 45 global exchanges.

• A tokenizer turns continuous OHLCV data into hierarchical discrete tokens, and an autoregressive Transformer is pre-trained on them.
• Models are released on Hugging Face; there are fine-tuning scripts and a 24-hour BTC/USDT forecast demo.

Turns one Claude Code session into a game dev studio

An agent and skill framework that organizes a Claude Code session like a game studio: directors, department leads, and specialists, each with defined responsibilities and escalation paths.

• 49 agents cover design, programming, art, audio, narrative, QA, and production.
• Skills are slash commands for each workflow; you still make every final decision.