Agents & tools

Translates your prompts to English before Claude reads them and streams translations of its replies back.

Mod that shows images pasted into Claude Code and lets you annotate them before sending.

Open-source Claude Code mod by arasovic. Its repo describes it as: Mods for Claude Code: function-hook plugins that add live panes and behaviour to the terminal UI. Shows what the community built in the first days of mods.

Open-source Claude Code mod by leopiney. Its repo describes it as: Voice coworker for Claude Code. Talk things through with a 3D wolf powered by ElevenLabs conversational AI; when you agree, it sends the prompt to Claude and speaks up when Claude is done. Shows what the…

A community Action that calls DeepSeek Harness to review pull requests, diagnose failing CI and turn issues into PRs. Brings the harness into everyday repo automation.

Translates the official Windows desktop app into Russian with sources, an installer and a video guide. Opens the harness to Russian-speaking users who could not read the UI.

Typed gateway client plus readers for the host's sand-data directory, built without inventing undocumented endpoints.

A group-chat agent framework from the NekroAI community that runs bots across chat platforms using DeepSeek Harness as the brain. Built for persona bots in community groups.

A pixel crab beside context and limit meters who gets nervous near a limit, levels up as you work and spends coins on hats.

A pixel pet that reacts to what Claude is doing, from chin-in-hand thinking to typing on a laptop and dozing off at night.

VM-less personal-agent infrastructure that deploys to Cloudflare Workers so products can ship a Muse-style assistant cheaply.

A reimplementation of the Cordis meta-framework on the JVM with revertible effects and reactive coeffects. Lets Java developers write the same plugin architecture DeepSeek Harness uses.

A prompt-bar mod for terminal and desktop with limit bars, a cache alert with sound, auto-compact at 99% and model/effort pickers.

Shows how long the session's prompt cache stays warm and can keep it alive while you are away so breaks do not cost a full re-cache.

An OpenAI-compatible Go gateway where Jev or a local Laya picks which LLM answers each prompt, with a decision benchmark.

Uses Clef to decide each step of an end-to-end test, shown with a shopping-flow demo.

Free open-source desktop app with AI teammates that each get their own computer, plus a video-understanding layer.

Local multi-agent workspace and swarm positioned as an open alternative to Dots, run on your own hardware.

A Rust server with custom Metal kernels that runs Laya in 15.5 ms on an M5 Mac, up to 4.1x faster than the reference server, behind the same API.

Progress bars beside the mode indicator that update whenever Claude Code re-measures your limits.

Live session cost, context window and plan-quota meters with Nerd Font icons, built for the mods API.

Shows pending channel messages, in-flight tools and subagents, rate limits and custom command cards in a one-line band and a side pane.

Clawd acts out what Claude is doing beside a small scene of your context, cache and limits, in the terminal and desktop app.

SGLang NEXTN setup at about 37 tok/s on code.