Claude Code · mods · code reference
The code reference for Claude Code mods.
Read real mods' code and start building from it. modscode reads public Claude Code mods on GitHub, gathers the code that does each thing you may want to build, and rebuilds what each mod draws from its code. For the terminal and the desktop app.
Read the code Browse the gallery
Check every mod yourself before you install it. Mods run unsandboxed with your permissions; modscode does not vouch for them.
From what you want to build to code
$.ui.openBand above the promptA strip right above the prompt: state or usage in a line or two.6 mods · on('ui.render', { component: 'AbovePrompt' })Status lineOne line under the prompt, with $.ui.status.6 mods · $.ui.status(text)ToastA short notice that goes away.6 mods · $.ui.toast(text, { timeoutMs })Deny a tool callBlock rm -rf or a force push. The mod form of PreToolUse.6 mods · on('tool.call', { tool: 'Bash' })Rewrite the promptFix the text before it is sent, or add a line only Claude reads.6 mods · on('prompt.submit')Add a slash commandA /command that runs your code.6 mods · $.command.registerKeep state with $.storeSave values across sessions.6 mods · $.store.getSettings (userConfig)Settings users can change, declared in plugin.json.6 mods · plugin.json userConfigAdd a tool for ClaudeA tool Claude can call, without an MCP server.6 mods · $.tool.registerExample: Band above the prompt · reviewer2 — In sessions opened at the Rev2Agent root, Reviewer 2 adds a one-line aside in a speech bubble above the prompt after each answer, beside an avatar whose expression fits it.
on('ui.render', { component: 'AbovePrompt' }, async ($, e, next) => {
const shown = bubble
const isPending = shown === null && pendingTurns.size > 0 && !e.props.isWorking
if ((shown === null && !isPending) || e.props.hasSurvey || e.props.view?.agentId !== undefined) {
return next(e)
}
let mine: RenderElement
if (e.surface === 'terminal') mine = drawTerminal($, e, shown)
else if (e.surface === 'desktop' || e.surface === 'vscode') mine = drawDesktop($, e as RenderInput<'AbovePrompt', 'desktop'>, shown)
else return next(e)
const { Box } = $.ui.resolve(e)
const below = await next(e)
return below ? Box({ flexDirection: 'column', children: [mine, below] }) : mine
})Most starred mods
Most GitHub stars first.
- 1blast-radiusSee what a risky command would change before it runs.deepseek-ai · Terminal + desktop, Pane, Band★ 244419
- 2replay-theaterStep through the last turn's edits, one diff at a time.deepseek-ai · Terminal + desktop, Pane, Band★ 244419
- 3ruview-liveRuView live sensing in Claude Code: /ruview opens a pane beside the transcript with three views: an overview of your CSI nodes (ESP32, Realtek) and 60 GHz radar (rates, loss, RSSI, presence, device-reported vitals, alerts), a live CSI ampl…ruvnet · Terminal + desktop, Pane, Status line★ 96675
- 4my-modA Claude Code mod scaffolded by ruflo-plugin-creator: counts the session's tool calls in a status line, adds /my-mod-status, and keeps a classic hook as the fallback where function hooks are off.ruvnet · Terminal + desktop, Status line★ 73986
- 5ruflo-adrADR lifecycle management — create, index, reconcile, supersede, check compliance, and link Architecture Decision Records to code via AgentDB hierarchical store + causal edges (supersedes/amends/depends-on/related)ruvnet · Hooks only★ 73986
- 6ruflo-agentAgent runtimes for ruflo — local WASM-sandboxed agents (rvagent: 10 wasm_agent_*/wasm_gallery_* MCP tools, built on @ruvector/rvagent-wasm + @ruvector/ruvllm-wasm per ADR-070) plus Anthropic Claude Managed Agents as a cloud backend (manage…ruvnet · Hooks only★ 73986
- 7ruflo-agentdbSubstrate plugin for Ruflo memory: AgentDB controller bridge (15 agentdb_* MCP tools), RuVector ONNX embeddings (10 embeddings_* tools incl. RaBitQ 32x quantization), and WASM HNSW pattern router (3 ruvllm_hnsw_* tools). As a mod (ADR-445)…ruvnet · Hooks only★ 73986
- 8ruflo-agntcyAGNTCY/Outshift Internet-of-Cognition runtime integration for Ruflo — SLIM secure transport for cross-host agent coordination, CASA intent-scoped tool authorization enforcement, optional IOC Layer 9 coordination events, and AGNTCY identity…ruvnet · Hooks only★ 73986
- 9ruflo-ai-teamGive Claude a tenant-isolated AI team that divides work, coordinates progress, recalls approved context, and returns evidence-backed results.ruvnet · Hooks only★ 73986
- 10ruflo-aidefenceAI safety scanning, PII detection, prompt injection defense, and adaptive threat learningruvnet · Hooks only★ 73986
From the gallery
All modsHow the code gets here
01
Found
About a thousand GitHub repositories are searched for mods with a hooks module.
02
Read
Its source is fetched as text at one commit and parsed, never run. What it draws is rebuilt from its JSX.
03
Cut out
From the listed mods, the code that does each thing (a pane, denying a tool call, a slash command...) is cut out with its licence.
04
Your check
Each page shows what the code reaches: network, programs, files, keys. Read it and decide before you install.
For building your own
Add the MCP connector and Claude can find real mods' code by pattern ("a pane with buttons", "deny rm -rf") and build from it, licence included.
https://modscode.com/mcpFAQ
What are Claude Code mods?
A mod is a Claude Code plugin made of JavaScript or TypeScript hooks that change how Claude Code looks and behaves. In the terminal and in the Code tab of the desktop app a mod can draw a pane beside the transcript, a band above the prompt, a status line and toasts, add slash commands, and deny or rewrite tool calls.
How do I learn to build a mod?
The code reference (/code/) has, for each pattern (pane, band, status line, toast, denying a tool call, rewriting the prompt, slash command, $.store, userConfig), the smallest form and real mods' code that does it. Run claude --plugin-dir . in your mod's folder to load it for one session without installing it.
How is a mod different from settings hooks (PreToolUse) or the statusLine setting?
Settings hooks and the statusLine setting run shell commands. A mod is JavaScript that runs inside Claude Code: it denies a tool call by returning { deny } from tool.call, shows a status line with $.ui.status, and can draw panes and buttons.
Which mods are listed?
Public mods on GitHub whose code is laid out so you can check it: every module read, all the code inside the mod, nothing minified, evaluated or encoded, and every call outside through $ in the open. The rules page has the details.
Is a listed mod safe?
modscode does not say so. Mods run unsandboxed with your permissions. Each page shows what the code reaches (network, programs, files, environment variables, keys in the source) for you to weigh; read the code and install the commit you read.
Do these mods work on Windows?
Claude Code in the terminal and the desktop app both run on Windows. A mod that starts a program Windows lacks (tail, date, open) is flagged on its page; that program will not run on Windows.
How is the design drawn without running the mod?
Its JSX is parsed with TypeScript's parser and the element tree Claude Code would get is rebuilt, with sample values where only the run time knows the text.
Can I use the code?
The smallest forms and starters are written by modscode; use them as you like. A mod's own code is shown under its licence, with the licence text, when the licence allows; otherwise it is linked. Keep the notice when you copy it.