@localcode/cli  ·  v3.1.0  ·  MIT open source

Code
offline.
Own everything.

A terminal AI coding assistant that runs 100% local with Ollama — or connect Claude, OpenAI, or Groq. 30+ slash commands, a plugin system, hooks, MCP servers, and Nyx, your ASCII co-pilot.

↓ Install free ⎇ GitHub
4
Providers
30+
Commands
100%
Free w/ Ollama
MIT
License
localcode — ~/projects/api
100% offline with Ollama Multi-provider — Claude · OpenAI · Groq · Ollama 30+ slash commands Plugin system — drop .js in ~/.localcode/plugins/ Hooks — PreToolUse · PostToolUse · Notification Memory files — .nyx.md global + project MCP servers — stdio + HTTP 4 built-in themes Vision input — /image with llava · GPT-4o · Claude TF-IDF codebase search with /index Session history — browse and restore MIT License — fork it, own it 100% offline with Ollama Multi-provider — Claude · OpenAI · Groq · Ollama 30+ slash commands Plugin system — drop .js in ~/.localcode/plugins/ Hooks — PreToolUse · PostToolUse · Notification Memory files — .nyx.md global + project MCP servers — stdio + HTTP 4 built-in themes Vision input — /image with llava · GPT-4o · Claude TF-IDF codebase search with /index Session history — browse and restore MIT License — fork it, own it
Quick start

Zero to first commit in 60 seconds.

One npm package. No config files, no daemons, no cloud setup required.

01

Install

npm install -g @localcode/cli
Node 18+ required. That's it.

02

Choose your provider

Run localcode and pick a provider. Choose ollama for fully free, no-key-needed local AI.

03

Chat & code

Ask anything. Nyx reads files, writes patches, runs shell commands — with permission prompts before every write.

04

Ship it

Run /commit to generate a conventional commit message. Nyx co-authors every one.

Features

Built different.

Everything Claude Code and Codex have — plus the things they'll never give you.

◉

100% local with Ollama

Run qwen2.5-coder, deepseek-r1, llama3.2, or any Ollama model. No API key. No cloud. Your code stays on your machine.

Free forever
⟳

Switch providers live

Use /provider claude mid-session. Switch from local Ollama to GPT-4o to Claude without restarting. Keys stored per-provider.

Multi-provider
⧉

Plugin system

Drop a .js file in ~/.localcode/plugins/ to add custom slash commands. Full access to the conversation context and tools.

Extensible
⚡

Hooks

PreToolUse, PostToolUse, and Notification hooks — run custom scripts before or after any tool call, just like Claude Code.

Automation
◈

Memory files

~/.nyx.md for global memory, .nyx.md in project root for project context. Both loaded automatically at startup.

Persistent context
⬡

MCP servers

Connect Model Context Protocol servers — both stdio and HTTP transports supported. Give the AI access to your databases, APIs, and tools.

Extensible
⊟

Vision input

/image path/to/screenshot.png — works with Claude, GPT-4o, and Ollama llava. Describe UI, debug layouts, analyze diagrams.

Multimodal
⎇

TF-IDF codebase search

/index builds a full semantic index of your project. Ask Nyx to find relevant files without passing your entire codebase as context.

Smart context
⚿

Permission prompts

Every file write, patch, or shell command asks first. Press y, n, or a (allow all). Full control, always.

Safety
Providers

Your model,
your choice.

Switch live with /provider <name>. API keys are stored per-provider and auto-loaded from env.

◉
Ollama qwen2.5-coder · llama3.2 · deepseek-r1 · mistral · gemma3 · llava Free · Fully local · No API key Runs 100% on your machine Requires Ollama running locally
✶
Claude claude-sonnet-4-5 · claude-opus-4-5 · claude-haiku-4-5 ANTHROPIC_API_KEY console.anthropic.com
▣
OpenAI gpt-4o · gpt-4o-mini · o1-mini · o3-mini OPENAI_API_KEY platform.openai.com/api-keys
◙
Groq llama-3.3-70b · mixtral-8x7b · gemma2-9b Free tier available console.groq.com/keys
Commands

30+ slash commands.
All in the terminal.

Type / to open the interactive picker. Every command has usage hints and tab completion.

/clearClear conversation history. Checkpoints are preserved.session
/compactSummarize and compress the conversation to save context window.session
/statusShow provider, model, key status, token count, and checkpoints.session
/checkpointSave a named snapshot. Usage: /checkpoint before-refactorsession
/restoreList or restore a saved checkpoint by ID.session
/historyBrowse and restore past sessions interactively.session
/themeSwitch UI theme: dark · nord · monokai · lightsession
/shareExport session as shareable markdown or HTML snippet.session
/pingTest connection to the current provider. Shows latency.session
/exitSave session state and quit the application.session
/theme
Four built-in themes: dark (default), nord, monokai, and light. Instant toggle, persisted across sessions.
❯ /theme nord
✓ Theme set to nord
/history
Browse your past sessions by date and topic. Select any session to restore its context — files, messages, and checkpoints.
/commitGenerate a conventional commit message from staged diff. Nyx co-authors.git
/gitRun git commands with AI assistance. /git status, /git log, etc.git
/diffShow all files modified by tool calls in this session.git
/reviewAI code review of a file or selection with actionable feedback.code
/testGenerate unit tests. Auto-detects Jest, Vitest, Mocha, pytest, etc.code
/explainExplain a file, function, or concept in plain English.code
/watchWatch a file for changes and auto-explain diffs in real time.code
/commit — AI-generated commits
Stage your changes, run /commit, and Nyx reads the diff and writes a conventional commit message. She's automatically added as co-author.
❯ /commit
⟳ Reading staged diff…
✓ feat(auth): add refresh token rotation
Co-authored-by: Nyx <nyx@thealxlabs.ca>
/contextInject a file or folder into the conversation. Usage: /context src/context
/indexBuild TF-IDF search index of your codebase. Enables smart context retrieval.context
/webFetch a URL and inject the page content as context.context
/imageSend an image to the model. Works with Claude, GPT-4o, and Ollama llava.context
/templateApply a saved prompt template. /template init, /template review, etc.context
/aliasCreate custom command aliases. /alias deploy="run npm run deploy"context
/index — TF-IDF codebase search
Build a full semantic index of your project. Instead of injecting thousands of lines of context, Nyx finds the 5-10 most relevant files automatically.
❯ /index
◈ Indexed 847 files in 1.2s
✓ TF-IDF index ready. Ask me anything.
/providerList or switch providers. Usage: /provider ollamaprovider
/modelSwitch model for the active provider. Usage: /model qwen2.5-coder:7bprovider
/apikeySet or update API key for current provider.provider
/pingTest provider connection and show latency.provider
/doctorRun diagnostics: check Ollama, API keys, Node version, and config.provider
/initRe-run the setup wizard. Pick provider, model, and configure keys.provider
/provider — live switching
Switch between any provider mid-session without losing context. Your API keys are stored per-provider and reloaded automatically.
❯ /provider claude
✓ Switched to claude-sonnet-4-5
❯ /provider ollama
✓ Switched to qwen2.5-coder:7b (local)
/allowallToggle permission prompts for all tool calls this session.tools
/templateManage and apply saved prompt templates.tools
/aliasCreate or list command aliases for common tasks.tools
/hookManage PreToolUse, PostToolUse, and Notification hooks.tools
/mcpConnect or list MCP (Model Context Protocol) servers.tools
/memoryView or edit the .nyx.md memory files (global and project).tools
/hook — automation
Run custom scripts before or after any tool call. Wire up linters, formatters, notifications, or anything else you want Nyx to trigger automatically.
// ~/.localcode/hooks/post-write.sh
npm run lint --fix $FILE
git add $FILE
Plugin system

Extend Nyx with
your own commands.

Drop a .js file in ~/.localcode/plugins/ and it becomes a first-class slash command.

// ~/.localcode/plugins/deploy.js
module.exports = {
  name: 'deploy',
  description: 'Deploy to production',
  async run({ context, tools }) {
    // Full access to AI context + tools
    const review = await context.ask(
      'Check for security issues'
    );
    if (review.safe) {
      await tools.shell('npm run deploy');
    }
  }
};
◈

Zero boilerplate

Name the file, export an object with name, description, and run. Nyx hot-reloads it on next command.

⚡

Full AI access

Plugins get the full conversation context, can ask the AI questions, and call any of LocalCode's built-in tools.

⧉

Share and publish

Package your plugin as an npm module. Anyone can install it with npm install -g localcode-plugin-deploy.

⎇

Hooks too

Plugins can register PreToolUse and PostToolUse hooks that fire automatically on every tool call.

Why LocalCode

The one they don't
want you to have.

Claude Code is great — but it locks you in. LocalCode gives you everything they have, plus the things that matter for serious developers.

Feature
LocalCode
Claude Code
Codex CLI
Fully local / offline mode
✓ Free with Ollama
✗ Always cloud
✗ Always cloud
Multi-provider support
✓ 4 providers
✗ Claude only
✗ OpenAI only
Switch provider live
✓ /provider cmd
✗
✗
Open source (MIT)
✓ Full source
✗ Proprietary
~ Partial
Plugin system
✓ Drop .js files
✗
✗
Hooks (Pre/PostToolUse)
✓
✓
✗
Memory files (.nyx.md)
✓ Global + project
✓ CLAUDE.md
✗
MCP server support
✓ stdio + HTTP
✓
✗
Vision / image input
✓ /image cmd
✓
✗
TF-IDF codebase search
✓ /index cmd
✗
✗
Session history / restore
✓ /history cmd
✗
✗
Free to use
✓ With Ollama
✗ Subscription
~ Free tier

Last updated March 2026. Subject to change as tools evolve.

/\_/\ ( ·.· ) > ♥ <
Your co-pilot

Meet Nyx.

Nyx is LocalCode's ASCII cat mascot. She lives in the header of every session, reacting in real time to what's happening — thinking when the model is working, lighting up green on success, turning red on errors.

She's not just decoration. Every git commit made through /commit carries her co-author signature — so she shows up as a real contributor on your GitHub repos.

Co-authored-by: Nyx <nyx@thealxlabs.ca>
Mood states
idle · thinking · happy · error · waiting
Co-author email
nyx@thealxlabs.ca
Appears in
Header · Commits · Errors · Setup wizard
Created by
TheAlxLabs · Toronto
Get started

Ship it to your
terminal now.

Node 18 or later required. Ollama is optional — but recommended for free local AI.

Global install (recommended)
npm install -g @localcode/cli
Installs the localcode command globally.
No install needed
npx @localcode/cli
Runs directly via npx. No global install required.
Then launch:
localcode

First run → setup wizard → pick provider → start coding.

Node.js 18+ macOS · Linux · WSL MIT License TypeScript source 0 telemetry