Jim McMillan Claim this profile

Specify a github or local repo, github pull request, arXiv or Sci-Hub paper, Youtube transcript or documentation URL on the web and scrape into a text file and clipboard for easier LLM ingestion

Kosmos: An AI Scientist for Autonomous Discovery - An implementation and adaptation to be driven by Claude Code or API - Based on the Kosmos AI Paper - https://arxiv.org/abs/2511.02824

Toolkit for measuring Claude Code and Codex performance over time against a baseline using SWEbench-lite dataset **No API key required for Max or Pro subscribers**

Claude Code Trace - Export Claude Code chat sessions into markdown and XML

Claude Code and Codex API Routing Proxy

Deterministic Python codebase analysis + LLM-powered deep investigation for AI coding assistants. Solves the cold start problem: Layer 1 extracts 37+ signals from AST/imports/git in seconds (95% token reduction). Layer 2 spawns parallel investigation agents producing comprehensive onboarding documents with file:line evidence citations.

Automated Document Intelligence Workflow

A full-featured, production-ready implementation that extends the core LangExtract library with Gemini 2.5 models, adding template management, batch processing, URL fetching, reference resolution, quality scoring, multi-pass extraction, and HTML/GIF visualization to create a complete document intelligence platform.

Command-line utility that crawls website docs and archives them in XML format for LLM ingestion

Comprehensive local repo report generation

Community Claude Code plugin marketplace - skills, agents, commands, and hooks for extending Claude Code. Import/export tools included.

Hey Claude - ultralight CLI launcher for Claude

Claude Code skill for spawning observable tmux sessions. Launch background Claude agents in visible terminal panes instead of blind subprocesses. Handles session management, command sending, and output capture. install_prompt.md users AskUserQuestion tool for user specific install customizations.

Autonomous coding agent right in your IDE, capable of creating/editing files, executing commands, and more with your permission every step of the way.

Notebooks for langchain

Prompts for future AI systems

The MCP Code Executor is an MCP server that allows LLMs to execute Python code within a specified Conda environment.

AI Email Analysis Pipeline with GPT-4, o1 Pro, qstash, digital_ocean, and resend

A Unix-native runtime for LLM agents. Agents interact via filesystem paths and CLI commands rather than framework-specific APIs.

convert screenshot in clipboard to base 64 and return to clipboard for pasting into remote terminal

Task Automation Orchestrator Agent (Tao Agent) using OpenAI Assistants API and function calling

Implementation of System 3 (meta-cognitive layer) from the Sophia paper as a Claude Code extension for persistent memory and self-improvement

A working example of RAG using LLama 2 70b and Llama Index

Fine-tune mistral-7B on 3090s, a100s, h100s

Claude and Gemini Code Review and Bugfix Competition Evaluation on 1.4 million token repo https://github.com/jimmc414/Kosmos

Automated detection of Federal Reserve language shifts in FOMC communications using keyword frequency analysis. Identifies semantic pivots, euphemism adoption, and narrative changes with 0-day detection lag. Supports both manual execution and fully autonomous operation via Claude Code agents.

Synthetic Data curation for post-training and structured data extraction

Toolkit for linearizing PDFs for LLM datasets/training

A modular graph-based Retrieval-Augmented Generation (RAG) system

ClaudeControl is a Python library built on pexpect that automatically discovers, fuzz tests, and automates any command-line program using pattern matching and session management.

Upload an XLSX or CSV to a pandas dataframe and display the top 10 results per dataframe

Efficient summarization of a Python codebase for LLM ingestion

prompt implementation inspired by https://news.ycombinator.com/item?id=42601847 which focuses on better code analysis techniques for AI

lightweight local voice recognition model to launch windows command strings, webpages or scripts

Legal Document Extraction Pipeline (Pydantic + Google Gemini 2.0 Flash) - PoC

Prompts and templates that let a stronger Claude model (Fable) build a deviation-resistant execution framework for a successor model (Opus) to run unattended

Monitors Claude Code documentation for changes. Fetches all pages, stores snapshots in SQLite, produces unified diffs, AI-powered change digests, and config audits.

OPAL: Orient-Plan-Act-Loop — a minimal agent harness with deterministic completion gates, designed from the LLM's perspective. Spec finalized, Phase 1/2 experiments complete.

TypeScript/JavaScript AST scanner for repo-xray — deterministic codebase analysis for AI coding assistants

Check domain availability via RDAP/WHOIS, generate name suggestions, monitor expiring domains, register via Porkbun API. Direct registry queries — no third-party services in the data path.

Automated trigger testing and closed-loop optimization for Claude Code skills. Uses claude -p or the Claude Agent SDK for inference. Measures precision/recall/F1, then rewrites frontmatter until the skill fires reliably.

A design pattern for Claude Code skills that separates conversation awareness from dense reference material. Routers (context: inherit) extract user intent; recipes stores (context: fork) hold unbounded reference content at zero token cost to the main context window. Includes 22 working example skill pairs.

CLI tool to manage Claude Code hooks with enable/disable toggling. Supports global and project-level settings with slash command integration.

The Power BI Desktop MCP Server is a tool that lets AI assistants like Claude interact with your Power BI models programmatically. It enables Claude to read your model structure, run DAX queries, create and modify measures, manage relationships, and perform advanced analytics - all through natural conversation.

Tools to build your own "taskmaster"

Collection of Claude Code workflow creation and visual design prompts

Bringing large-language models and chat to web browsers. Everything runs inside the browser with no server support.

Scan food labels and barcodes to identify additives using on-device OCR with a signed SQLite additive database for personalized dietary and regulatory risk explanations.

A Python terminal application that administers an interactive MBTI personality test with 88 questions, featuring session persistence, animated UI, and detailed personality type analysis with export capabilities.

Convert PDF to markdown quickly with high accuracy

OCR a folder optionally through Google Vision API, AWS Textract API or Microsoft Azure Computer Vision API

R2R is a RAG (Retrieval-Augmented Generation) engine with a RESTful API and prod features. Including hybrid search, knowledge graphs, and more..

Python implementation of Lewis Carrols method of determining the day of the week

Instructions, source code, and misc. resources needed for building a Tiny ML-powered artificial nose.

Open-Source Data Platform 🐰 — Directus wraps any SQL database with a real-time GraphQL+REST API and an intuitive app for non-technical users.

In this paper, we apply machine learning to create bots for Halite III, @twosigma's annual A.I. competition. We develop one classifier using Support Vector Machine with Supervised Learning, and one using a Deep Neural Network with Reinforcement Learning

Identify the face of the owner and open the locked door. This is made using openCv, numpy and operated using Arduino.

📜 A compendium of CLI commands I can't stop looking up

This program is a text proofreading and punctuation fixing tool that uses OpenAI's ChatGPT API. The user is prompted to speak their input, which is then converted to text and sent to the API. The API processes the input and returns a response, which is then copied to the clipboard. The program plays a sound to indicate success.

OpenAI-powered semantic search for any YouTube playlist – featuring the All-In Podcast. 💪

Create a compiler listing with IGYCRCTL

Ask the user to supply a directory and all supported image files in the directory are OCR'd and there is a .txt file created contains the same name as the image filename.

Official repo for the paper: Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models

LLM Chain for answering questions from documents with citations

The ChatGPT Retrieval Plugin lets you easily search and find personal or work documents by asking questions in everyday language.

Chat with your notes in Obsidian! Plus, see what's most relevant in real-time! Interact and stay organized. Powered by OpenAI ChatGPT, GPT-4 & Embeddings.

Specify what you want it to build, the AI asks for clarification, and then builds it.

Game for a 9 year old to practice multiplication tables

A simple tool for visually comparing two PDF files

Super simple MLX (apple silicon) CLIP based photo similarity web app

A framework for Claude Opus to intelligently orchestrate subagents.

Infinity is a high-throughput, low-latency REST API for serving vector embeddings, supporting a wide range of text-embedding models and frameworks.

⚗️ distilabel is a framework for synthetic data and AI feedback for AI engineers that require high-quality outputs, full data ownership, and overall efficiency.

Creates a new Github repo and commits the current directory into it.

v2 of Task Automation Orchestrator (Tao) that removes AI for a more traditional and deterministic approach

OpenAI's Code Interpreter in your terminal, running locally

Daily Interest Calculator

Matrix Code Display in Python

ADHD Toolkit CBT-Based Worksheets for Better Focus and Control

Automatic Evals for LLMs

Reasoning Augmented Generation

Lightweight coding agent that runs in your terminal

Human in the Loop Automation Manager "Do nothing script"

news aggregator

Claude Engineer is an interactive command-line interface (CLI) that leverages the power of Anthropic's Claude-3.5-Sonnet model to assist with software development tasks. This tool combines the capabilities of a large language model with practical file system operations and web search functionality.

A react-based starter app for using the Live API over websockets with Gemini

Gemini 2.5 Pro code assistant

An Exploration of Design Patterns for Claude CLI (`claude -p ""`) Orchestration

The open-source RAG platform: built-in citations, deep research, 22+ file formats, partitions, MCP server, and more.

Present the days tasks based on current date and task schedules

Simple CLI Recurring Deadline Reminder

No description.

Get alerted the instant Anthropic's Claude Fable 5 comes back online after its government-ordered suspension. Watches via your Claude Max subscription and notifies by terminal, desktop, email, and phone (ntfy). Stdlib-only, cross-platform.

Constituent pressure scoring and contact tool for the US Senate Iran war funding vote. Ranks senators by expected impact, routes you to a phone call in under 5 minutes. No backend, no accounts.

Claude Code metadata extraction and visualization project

Local-first GTO poker solver — Rust CFR engine, TypeScript/React frontend, Tauri v2. Complete build specification: 195 requirements, 8 crates, 8 milestones. Research-backed by comprehensive GTO Wizard reverse-engineering study.

Testing of `claude mcp serve`

Bias-free college football rankings using iterative convergence. Derives team ratings from game outcomes and opponent strength, not human voting.

Reverse-engineer legacy system behavior through controlled database experimentation using snapshot-diff methodology

These are commands I use with agents, mostly Claude

Claude-first task management CLI with AI-assisted development workflows

Python task orchestration framework with validation-based error handling for Windows Task Scheduler automation

BioReason: Incentivizing Multimodal Biological Reasoning within a DNA-LLM Model | NeurIPS '25

An Optimized Speech-to-Text Pipeline for the Whisper Model Supporting Multiple Inference Engine

Proof of concept: Block Claude Code's built-in AskUserQuestion tool with a PreToolUse hook and redirect to a custom MCP tool via permissionDecisionReason.

Lightweight dependency tracking to prevent documentation drift in AI-assisted development workflows. Includes CLAUDE.md protocol, pre-commit hook, and GitHub Action.

No description.

PII Masker is an open-source tool for protecting sensitive data by automatically detecting and masking PII using advanced AI, powered by DeBERTa-v3. It provides high-precision detection, scalable performance, and a simple Python API for seamless integration into workflows, ensuring privacy compliance in various industries.

Official code for NeurIPS 2025 paper "AutoDiscovery: Open-ended Scientific Discovery via Bayesian Surprise"

Capturing feedback for https://stardev.io

Claude Code skills for coordinating multiple parallel sessions on the same codebase using git worktrees as the communication layer.

A Python CLI tool for benchmarking LLM logical deduction capabilities through configurable Mastermind games. Supports multiple models via LiteLLM and manual clipboard mode for web UIs, with comprehensive JSONL logging for analysis.

R&D directory for the Kosmos project

An open-source AI agent that brings the power of Gemini directly into your terminal.

An end-to-end Data Scientist

An agentic Machine Learning Engineer

A general purpose scientific writer

A set of ready to use scientific skills for Claude

MCP server for searching and retrieving Claude Agent Skills using vector search

Claude Code proxy for Microsoft Data Formulator - handles inference and drives UI automation

Research-driven framework for detecting and resolving prompt conflicts that cause LLM output corruption. Based on 6 peer-reviewed papers on superposition interference and emergent misalignment. Provides meta-prompts analyzing 15 conflict categories with automated disambiguation.

Creates a local IBM mainframe environment sandbox (MVS 3.8J) controllable via AI Copilot (Claude Code or Codex), allowing mainframe operations, & run CICS transactions. The AI Copilot and TN3270 automation demonstrated here can connect directly to live production IBM mainframes, providing identical AI control for enterprise mainframe operations.

Claude_CLI_Control is a Python library built on pexpect that autonomously discovers, fuzz tests, automates, records and replays any CLI program using Claude Code with zero configuration

documents used for claude_control (codex) vs claude_cli_control (claude_code) brownfield test

This project is a PoC for a Claude Code and Codex collaborative project with github actions

An Advanced Vector-Embedding Operations Toolkit and Cookbook

A Git integrated semantic merge engine that operates on typed ASTs and symbol graphs using CRDT ordered lists to automerge parallel refactors with deterministic, formatter preserving, compiler verified results.

converts long Slack, Teams or Outlook threads into concise auditable briefs that list decisions risks actions owners and links back to the exact messages.

vendor adaptive invoice screener that learns each vendor’s own patterns for totals, line items, invoice numbers, and bank accounts so it can detect duplicates and anomalies with context instead of applying a single global standard

a discussion between a human and AI

A MCP server for our beloved terminal multiplexer tmux.

Get your documents ready for gen AI

aider is AI pair programming in your terminal

Utility to verify specific files have been refreshed according to defined schedule

Historical S&P 500 Monthly Investment Growth Calculator

No description.

Demo showing how to use the OpenAI Realtime API to navigate a 3D scene via tool calling

Execute Python and Capture Errors in Clipboard for pasting into an LLM

PDF to Markdown with vision models

🖼️ A tool for efficient processing of spectral images with Python.

Mirror test LLM eval for Dec 2024 frontier models

Minimalist **console-based adaptation** of the classic Flappy Bird game, rendered in text

extract emails into json by date, from, subject and body search criteria

scheduler_implementation

subprocess_orchestrator

turnkey self-hosted offline transcription and diarization service with llm summary

Dolphin System Messages

No description.

User-friendly WebUI for LLMs (Formerly Ollama WebUI)

A file management automation tool with SQL-like syntax.

No description.

Interact privately with your documents using the power of GPT, 100% privately, no data leaks

Python script to quickly benchmark the performance of IO, CPU, and memory operations on a computer and generate a score for comparison

JAX implementation of OpenAI's Whisper model for up to 70x speed-up on TPU.

Simple program that just evaluates whether file1 is identical to file2 or not, and if it's not, returns the differences.

Command line Python utility to convert a YouTube link to an mp3 file

GUI editor for .ini config files using wxPython and ConfigObj

Input: LLMs\nOutput: <json>😽</json>

PipMyRide is a Python program that generates a pip install command from remote or local repositories.

Chat with a CSV

An open source ChatGPT UI.

No description.

An experimental open-source attempt to make GPT-4 fully autonomous.

No description.

e2b (english2bits) is an IDE powered by AI agents. Developers describe what they want to build by writing documentation. Then let AI agents with access to tools do the coding work.

No description.

An AI-powered tool for automating literature reviews with concise summaries and relevant citations.

No description.

Fork of Langchain Repo as of April 2023

No description.

No description.

Use ChatGPT to summarize the arXiv papers.

Scan a pdf and create a distinct list of the words in the document ordered in descending number of occurrence

Python prototype for UPC scanning Android and Iphone App that allows the tracking of food consumption via UPC cross referenced with symptoms of food reaction or intolerance and produce report showing correlations

Speech to text input concatenated with the text currently in the clipboard, and sends the combined text to the OpenAI ChatGPT API to generate a response. The response is then placed in the clipboard and a ding sound is played to indicate success.

Create a distinct list of words contained on a webpage and subpages

No description.

Parse cash receipt journal to Excel

Single file PHP script that adds a REST API to a SQL database

Config files for my GitHub profile.