Claude Code developer tools

Method, repo template and VS Code extension for keeping the reasoning from human-AI work in Git: your agent turns your notes into Architectural Deliberation Records and asks open questions with options, a recommendation, rationale, confidence and a falsifier, which you accept or override; the extension's MCP server lets agents query the record.

Open-source (MIT) relay that lets coding sessions from different vendors work together, tested with Claude Code and Codex: one session leads, the other follows, the relay passes each turn across, records it on your disk and stops when something needs your decision, with a web console to watch and answer. macOS, Python 3.11+.

Claude Code mod that reads your plan's 5-hour and weekly usage and holds subagent fan-out, heavy-context prompts and retry loops before they spend your window, asking in Claude Code's own dialog. Local only, no network; MIT licensed.

Open-source (MIT) terminal UI that draws your Claude Code sessions and their forks as a tree or flowchart: move between parents and children, filter to this directory or the last 24 hours, and resume any session with claude --resume. It infers forks from shared message IDs because Claude Code records no parent link. Python 3.11+, pipx install arvre.

Claude Code plugin that warns Claude at 50% context so it finishes its current piece, writes itself a handoff note and compacts at a clean break instead of mid-task. MIT-licensed.

iPhone app plus a small daemon that puts Claude Code and Codex sessions from your Macs and Linux machines on one screen: see every agent waiting on you ranked by risk, approve or deny the exact command (risky ones need a hold and Face ID), answer questions, review diffs and take over the live terminal. Connects over your own Tailscale tailnet with no CodeConnect server or account. The app is free; iOS 17+, macOS 13+ or Linux x86_64/arm64.

Open-source (MIT) boilerplate for running Claude Code as a software factory: a sandboxed container, a deterministic bash-command regulator, lint hooks for custom rules and a multi-agent workflow generator.

Opaline

Analytics & Monitoring

Team-wide, message-level analytics for Claude Code and Codex sessions: token cost, time and skill usage for every message across your team; open-source CLI, paid team plans.

Open-source (Apache-2.0) single Go binary that puts your Claude Code, Codex and OpenCode sessions on one local IRC channel with you, so they tell each other when something changes and ask when they need input; you read every line in any IRC client, only your nick carries authority, and a breaker stops runaway loops until you write.

Open-source (MIT) CLI, hooks and skill that give every Claude Code or Codex session a unique name and emoji drafted from a roster you pick, so notifications and messages are traceable and agents can refer to each other; rostr stats shows who is working on what. Works with tmux and herdr; local by default, optional end-to-end encrypted cloud roster.

Runs Claude Code or OpenAI Codex CLI unattended on a brief, fenced to the folders you give it and under a per-run dollar budget enforced by a loopback API proxy, leaving one ledger line per run. Single Python file, standard library only, macOS only, Apache-2.0.

Open-source (MIT) Claude Code plugin that notices tasks and instructions you keep repeating and offers to save them as a /skill or a CLAUDE.md rule; data stays in a local SQLite file, no telemetry, and it never creates a skill without you.

Searchable library of engineering skills behind an MCP server: Claude Code and Codex find and load the right expert instructions for the task as they work, and teams can keep a shared private skill library.

Open-source (Apache-2.0) API key leak detection for AI agents: plants fake keys where real ones live, runs your agent's flows and reports whether they reached traces, logs, console output, the model or other services (leak, clean or incomplete), and in Claude Code shows which of your real keys entered or left a session, with an optional hook that redacts them. pip install weir-scan.