v0.2.7 — TUI Crash Fix

Fix

• Fixed terminal crash when using /models, /provider:add, or /settings commands
• Root cause: bold: true is not a valid style property in @OpenTui <text> elements, causing the terminal renderer to crash
• Also disabled chat keyboard handler when provider modals are open to prevent input conflicts

Full Changelog: v0.2.6...v0.2.7

v0.2.6 — Universal Multi-Provider System

What's New

Universal Multi-Provider System

Connect any LLM provider — 35+ providers supported out of the box across 9 categories:

Category	Providers
Major Paid	Anthropic, OpenAI, Google Gemini, xAI (Grok)
Aggregators	OpenRouter, Groq, Together AI, Fireworks AI, AIHubMix, 302.AI
Specialized	Mistral, DeepSeek, Cohere, Perplexity, Replicate
Chinese	Alibaba/Qwen, Moonshot, Xiaomi, Z.AI
Enterprise	AWS Bedrock, Azure OpenAI, GitHub Models
Free/Local	Ollama, LM Studio, Ollama Cloud
GPU Cloud	Nvidia, Vultr, Cerebras, DeepInfra
Coding Tools	OpenCode Zen, KiloCode, CodeBuff

New Slash Commands

• /provider:add — Interactive wizard to add any provider
• /provider:list — List configured providers with status
• /provider:remove <id> — Remove a provider
• /provider:test <id> — Test connection with latency
• /model:list — Searchable model picker with provider filtering
• /model:set <provider/model> — Set active model
• /model:info — Show current model details
• /settings — Open provider & model settings panel

Architecture

• Config: ~/.config/levelcode/providers.json — persistent provider configuration
• Catalog: Fetches from models.dev/api.json with local caching
• Auto-detect: Scans for Ollama and LM Studio on localhost
• Routing: Integrated into getModelForRequest() — providers.json checked before legacy routing
• TUI: Provider wizard, model picker, and settings panel components

Bug Fixes

• Fixed all pre-existing TypeScript type errors across the monorepo (0 errors)
• Fixed TeamTask priority field missing from Zod schema
• Fixed JSONValue type mismatches in agent-runtime tool handlers
• Fixed it.each() usage incompatible with Bun test runner
• Fixed CLI minor type issues (border styles, module declarations)
• Fixed SDK test type mismatches

---

Full Changelog: v0.2.5...v0.2.6

LevelCode v0.0.9 - Standalone Mode

LevelCode v0.0.9 - Standalone Mode 🚀

LevelCode now works fully standalone — no backend server required. Just set your API key and go.

What's New

Standalone Mode (No Backend Required)

• Direct OpenRouter support — Set OPENROUTER_API_KEY to access 200+ models directly
• Direct Anthropic support — Set ANTHROPIC_API_KEY for Claude models directly
• No login required — Skip the web login flow entirely
• No credits/billing — Unlimited usage in standalone mode
• Zero configuration — npx @levelcode/cli just works

Removed Legacy SaaS Features

• Referral system removed (not needed for open-source)
• Credits/billing tracking removed (unlimited in standalone)
• Out-of-credits blocking removed
• Agent publishing disabled in standalone mode
• Backend connection polling disabled

Kept Intact

• Ads system — Gracefully degrades when backend unavailable, helps fund the open-source project
• Claude OAuth (BYOK) — Connect your Claude Pro/Max subscription for direct access
• Multi-agent architecture — File Picker, Planner, Editor, Reviewer all work locally
• All 200+ models via OpenRouter

Quick Start

# Install
npm install -g @levelcode/cli

# Set your API key (pick one)
export OPENROUTER_API_KEY=sk-or-...    # For 200+ models
# OR
export ANTHROPIC_API_KEY=sk-ant-...     # For Claude directly

# Run
levelcode

Packages

Package	Version
@levelcode/cli	0.0.9
@levelcode/sdk	0.0.3

Full Changelog

v0.0.6...v0.0.9

LevelCode v0.0.001 - Initial Release

LevelCode v0.0.001 - Initial Release

The first release of LevelCode - an open-source AI coding agent that outperforms Claude Code.

Features

• Multi-Agent Architecture: Specialized agents (File Picker, Planner, Editor, Reviewer) work together for precise codebase editing
• 200+ Model Support: Use any model via OpenRouter - Claude, GPT, DeepSeek, Qwen, Gemini, Llama, and more
• TypeScript SDK: Build LevelCode into your applications programmatically
• Terminal-First CLI: Beautiful, responsive terminal UI with real-time streaming
• Custom Agent Workflows: Create specialized agents with TypeScript generators
• Evaluation Benchmarks: 175+ real-world coding tasks (61% success rate vs Claude Code's 53%)

Installation

# CLI
npm install -g levelcode

# SDK
npm install @levelcode/sdk

Quick Start

export OPENROUTER_API_KEY="your-key"
cd your-project
levelcode

Documentation

• Installation Guide
• Configuration
• SDK Reference
• Custom Agents

Credits

Created by Yethikrishna R

---

Full Changelog: https://github.com/yethikrishna/levelcode/commits/v0.0.001