Claude-Code

Claude’s verbose output — long caveat lists, hedging, verification dumps, and “token vomit” — wastes paid tokens and human reading time every time you copy a reply into a PR, doc, or commit. Vomit fixes this by piping Claude’s message through a separate, local LLM that rewrites it into concise prose before it ever reaches your screen. The cleanup pass runs on your own machine for zero hosted-token cost, and measured over 100 agent iterations it cut downstream token waste by roughly 20%. This guide shows you the root cause, how to set up Vomit, and why separating the reasoning model from the style model is a pattern worth stealing. ...

Frugal Tokens is a free, local, read-only web dashboard that shows exactly how many tokens each of your AI coding agent sessions consumed and what those tokens cost, across OpenCode, Claude Code, PI, Codex, and Cursor. It runs entirely on your machine, never mutates your agent databases, and surfaces hidden costs like prompt-cache misses that most developers never see. This guide walks you through installing it, pointing it at your agents, and reading the per-session explorer to cut your coding agent spend. ...

A local-first control plane agent like Swarm watches every Claude Code session on your machine live so you can see each agent’s tool calls, reasoning, token spend, and cost from one dashboard instead of cycling through terminal tabs. It keeps a ledger of who holds which task and worktree, turns CLAUDE.md “never do X” prose into real permission decisions, and runs entirely offline with no account and no telemetry. What Is a Local-First Control Plane for Coding Agents? A control plane is the management layer that decides how the agents underneath it behave. In IT operations, the control plane and the data plane are separated: the data plane does the work, while the control plane watches, meters, and governs. ra3orblade/swarm applies that same separation to AI coding agents, with one crucial twist: the entire control plane runs on your own machine. ...

agent-shell is a native Emacs shell that lets you talk to any AI coding agent from a single buffer, powered by ACP (Agent Client Protocol) rather than any single vendor. It supports 18+ agents including Claude Agent, Codex, Gemini CLI, Goose, and Grok Build, and by September 2026 it has grown to roughly 1,824 GitHub stars. You install it from MELPA, configure authentication with environment variables, and switch agents without leaving Emacs. ...

Portable Handoff: How to Keep Context Between AI Chats, Models, and Coding Agents AI coding agents start every session with a fresh context window, so anything you do not persist is re-learned from scratch. A portable handoff file solves this: you write your instructions, preferences, and project conventions once, then symlink or import that single source of truth into every agent you use — Claude, Gemini, Cursor, or Codex — so context follows you across chats, models, and tools without drifting. ...

ARC-Skill is a lightweight agent skill that lets an unmodified Claude Code agent finish all 25 ARC-AGI-3 games and all 183 levels at a perfect 100.00 RHAE score, using just 7,645 actions versus a median human’s 17,135. It works by enforcing a single rule: before the agent presses any button, it must write down exactly what that press will do to the grid, and the harness refuses to act until a prediction is recorded and graded. This article explains how that one rule unlocks frontier-level performance, how ARC-Skill compares to heavier harness systems, and what it teaches about agentic skill engineering. ...

Anthropic is quietly A/B testing reduced default effort levels in Claude Code, and the change does not make the coding agent dumber — it makes it less proactive. Effort controls how much autonomous work Claude performs per turn (reading files, running tests, double-checking its own output) before responding or asking for context. When the default drops, you get faster, cheaper turns that skip deep investigation, which is fine for scoped tasks but can silently degrade complex multi-file refactors. ...

Frugal Tokens is a free, local-first, read-only tool that lets you see exactly how many tokens each of your AI coding agents consumed and what that cost you, across OpenCode, Claude Code, PI, Codex, and Cursor. It runs locally in Deno, never touches your agent databases, and surfaces per-session token usage plus the hidden cost of broken prompt-cache reuse. It is the sharpest per-session cost diagnostic we tested for teams that want to measure ROI on AI coding spend. ...

Open-source memory for coding agents synced over SSH lets you carry your AI assistant’s context, preferences, and project knowledge across every machine you work on — without a cloud service. The most portable approach is a git-backed markdown store: symlink your agent’s memory files, push and pull over SSH, and let auto-sync hooks handle the rest. This guide walks through the three proven methods and how to pick the right one. ...

Heimdall is an open-source, CPU-only knowledge layer that gives AI coding agents a verified, self-healing memory across every repository you work in. Instead of returning plausible-but-unverified matches, every search result carries a trust verdict — STRONG, WEAK, REBUILT, or STALE — re-checked against your live filesystem at query time. It indexes with tree-sitter and local embeddings, spends zero tokens on memory maintenance, and re-anchors moved files automatically. What is Heimdall and why does it exist? Heimdall is a trust-verified knowledge layer built specifically for AI coding agents. It was released on Hacker News on August 22, 2026, and is published on npm as @ariantdeva/heimdall at version 0.2.1. The project’s core claim is simple: most agent memory tools return results that look right but have not been checked against the actual state of your codebase. Heimdall exists to close that gap. ...