Open-source and side work.

chiron

A mentor kit for coding agents, named after the centaur who trained heroes. When you fall back from a top-tier model to a weaker one, the weaker one inherits the strong one's ways of working: an operating doctrine (verify before acting, no workarounds, honest completion claims) plus enforcement hooks for Claude Code and Codex CLI that mechanically block check-bypassing, test deletion, and unverified "done" claims. Includes an A/B eval harness with pre-registered checks that measures whether the kit actually changes model behavior, and publishes the numbers either way. Install by pointing your coding agent at the repo's INSTALL-PROMPT.md; it clones, wires, and self-tests.

claude-can-speak

Lets Claude Code decide what to say out loud. Most "speak Claude aloud" tools read every reply at you; this one leads with a speak skill that lets Claude voice only what is worth hearing (a "build is done", a heads-up while you look away, a shoutout), while everything else stays text. A speak-everything firehose (its own on/off switch, off by default) is included too, but the deliberate skill is the point. Local neural text-to-speech (Kokoro for natural English, Piper for German, Turkish, and more) runs in Docker, so nothing goes to the cloud. Install with npm install -g claude-can-speak.

edgecall

Route a plain-language request to one of your own functions, using a deliberately weak local model. A small self-hosted, OpenAI-compatible API: you POST a sentence and a tiny model (the kind that runs on a Raspberry Pi or a phone) reads a short paginated menu of your registered functions, picks one, and edgecall runs it and returns the result. The model never sees the whole list at once; it pages through a few options at a time with a "next" choice, so all a weak model has to do is pick one item off a short menu. No embedding model, no index to maintain. Add a function by dropping in a Python file with an id and a one-line description. Talks to any OpenAI-compatible /v1 endpoint.

doublethink

ntfy you can actually trust: a publish/subscribe message broker as easy to stand up as ntfy, but with genuinely private channels. ntfy is effortless because its topic name is the only gate, which is also why its topics are effectively public. doublethink keeps the minutes-to-set-up ergonomics and adds real privacy: creating a private channel is self-service (one request, no operator) and returns a high-entropy shared secret. That secret is the gate, whoever holds it can join the channel and no one else can, and the broker never sees it, so messages are end-to-end encrypted between the parties who share the secret and the operator cannot read them. Knowing a channel name grants no access. Bidirectional, asynchronous, streaming, with opt-in plaintext public topics too. Cross-platform: run it in Docker or as a single native binary. There is also a sideloadable Android app that subscribes to topics and notifies you the moment a message arrives.

agent-doublethink

Keep two coding agents on two machines working the same project in sync, over an end-to-end-encrypted channel the broker cannot read. A small, agent-agnostic MCP server: point Claude Code, Codex, or Gemini at it and a partner agent on another machine at the same channel, and the two exchange plans, decisions, blockers, and divergence flags as they work. Built on doublethink, so the broker only ever sees ciphertext. Roles are auto-discovered; a one-line digest of your partner's state surfaces at the start of each turn so you catch drift early.

portunus-ovpn

Native KDE Plasma 6 panel widget for switching OpenVPN connections. Import, edit, remove and toggle .ovpn profiles from the panel; one VPN active at a time by default. Drives NetworkManager and KWallet directly, no second VPN daemon and no third-party PPA. Named after Portunus, the Roman god of keys and gates.

vigil

How to build an always-on, autonomous AI operations assistant, and how to keep it diligent. A documentation project: the architecture of a persistent AI operator (per-invocation agent subprocess with a resumable session, supervisor, transport split, channel multiplexers, MCP tools) as a reference design, plus the load-bearing piece most write-ups skip: the doctrine + hooks pattern that re-asserts an engineering standard on every turn so an auto-approved agent keeps verifying before it acts, refuses workarounds, and never claims done without a real run.

hydra-rag-hooks

apt install, then type rag <question> in either Claude Code or OpenAI's Codex CLI. UserPromptSubmit hooks for both CLIs from one package. The hook auto-indexes the project on first use, then prepends retrieved chunks from a local LanceDB index before the model sees the prompt. Both CLIs share the same .hydra-index/ store and the same bundled MCP server for model-decided retrieval. Successor to claude-rag-hook.

lillycoder

A local-first coder REPL with file and shell tools. Drop into any folder, bring your own LLM, ship code without leaving the terminal.

hydra-llm

Run local LLMs the easy way. Pick a model, download it, chat with it. No cloud, no API key, no telemetry.

hydracoder

Enter a project goal and local models plan it, build it, and review it, with no hosted AI in the loop. A dark web UI shows each model as a live terminal. Built on hydra-llm and lillycoder, so it runs entirely on your own hardware.

inhibit-charge

Park your Linux laptop battery at a target charge. AC powers the system, the battery sits idle. No trickle cycling.

meowtrics

Animated tray emoji that gossips about your machine. Sweats when the CPU is hot, naps when you are idle.

herald

A small Linux daemon and CLI that prints a quote at the top of every new terminal and at login.

baseline-engine

Browser-native game engine with built-in auth, multiplayer, and editor in one stack.

tekir

Self-explanatory HTTP responses for AI agents. An RFC 9457 extension that lets a service tell an agent why a request failed and what to try next, in a way the agent can actually parse.

clauding

Autonomous AI dev environment that bundles Claude, Codex, and Gemini into one Docker stack. One command, one workspace, three agents.

pdf-weaver

Styled and branded PDF generation in high volumes. Built for mass invoicing, recipes, certificates, and any templated document workflow.

genuine-captcha

Privacy-focused CAPTCHA solution. Protect forms without compromising user privacy or experience.

sftp2redis-bridge

Bridge system that ingests files from SFTP into Redis streams for downstream processing.

kartakip

Snow coverage monitoring for a specific region in Turkey, built on Sentinel-2 imagery and NDSI analysis.

leafwallet

A paradigm for accountless crypto wallet generation and ownership. An Ethereum dApp library for wallet-free transaction signing.

caleidoscode