Recently Evaluated

Graphite’s Five Percent research article analyzes writing-style differences in Claude Opus 5.5, compares it with prior Claude versions and GPT-6 Astra, and includes charts, tell counts, and downloadable data.

A GitHub repo for checking recorded agent runs against a TLA+ specification of the surrounding protocol. It models retries, acknowledgements, approvals, budgets, and can validate JSONL runs against the same spec.

A Claude Code skill and plugin that bulk-deletes old scheduled-task runs while keeping the newest N and preserving runs where a user replied.

An arXiv paper describing a governed, local-first memory system for AI agents, with audit trails, role-based access, verified erasure, and fault-injection results. The page also links to associated code and a Zenodo DOI.

Self-hosted cloud coding agent infrastructure for running Claude Code and Codex in per-thread sandboxes with a single Docker Compose setup.

Plannotator documents a workflow for letting AI agents ask questions in markdown-style question blocks and collecting answers asynchronously in one place. The page explains how it works, where it works, and how agents and reviewers exchange responses.

An essay about why AI agents should use project documentation instead of conversation memory, with a description of the author’s Operator Memory system for markdown-based agent memory.

Television is a visual interface for personal agents that helps users and agents create, arrange, and reuse artifacts, with channels for organizing work and reusable skills for common interface patterns.

A tiny programming language for AI agents that bakes in permissions, bounded retries, verification, and a tamper-evident audit log.

An MCP server that converts public URLs into clean Markdown for AI agents and MCP-compatible clients. The repo includes setup instructions, a usage example, a direct API option, and notes on supported clients like Claude Desktop and Cursor.

A repo for keeping coding agents from overengineering by injecting a ruleset during coding and running a slop-check after edits. It supports multiple agent setups, including Claude Code, Codex, OpenCode, Cursor, Copilot, and generic AGENTS.md-based workflows.

Browser-based RAG app that runs entirely client-side, using BM25/Lunr.js for retrieval and direct LLM provider APIs for answering questions without a backend.

A benchmark page for frontier models racing to write a StarCraft: Brood War bot in C++ using BWAPI and OpenBW, climbing tiers of increasingly strong Protoss opponents.

A lightweight Rust terminal notes app with SQLite full-text search and an MCP server so Claude can read, create, and update notes.

A tool that lets AI agents control Figma through a dedicated Chromium session and CDP, with MCP and HTTP API access, install steps, and agent/client setup guidance.

A deterministic security scanner and adversarial flight simulator for AI agents, built for red-teaming LLM workflows and MCP-compatible setups.

A command-line tool that sends fake RFC 3464 bounce messages to make spam senders think an address does not exist. The README explains the workflow, configuration, batch mode, multiple-account handling, and provider-specific caveats.

Portus is a Rust-based API and AI gateway for Kubernetes that handles HTTP, LLM, and MCP traffic on one data plane. The page shows Gateway API conformance, performance benchmarks, AI key and budget controls, and MCP routing examples.

A blog post documenting an agent-assisted prototype of a Choplifter-style game for Apple Vision Pro, including gameplay basics, Blender asset creation, and the author’s workflow using Astra and Claude Opus 5.5.

A self-hosted proxy and router for Claude Code that lets teams pool multiple Claude accounts and choose which account serves a session.

A data report comparing hospital list prices with negotiated insurer rates across 482 U.S. hospitals, with service-level examples and a hospital lookup section.

Premortem is an AI red-teaming tool for startup ideas. Users paste an idea, six specialized agents critique it, and a seventh agent synthesizes the findings into a memo.

Local tools for agentic software development: a Rust and egui workspace that groups tasks, shells, coding agents, and code review in one place, with remote browser access and reusable UI/terminal components.

Spens is a developer tool for running coding agents in sandboxed Docker containers with network controls, secret handling, and detailed logs of LLM calls, tool use, and file changes.