This is a community-enhanced fork of 21st.dev/1Code
A powerful, locally-focused coding agent client with enhanced features for privacy-conscious developers and advanced model management.
Fork Version: v0.0.90 (upstream: v0.0.72)
This fork extends the original 1Code with 48+ unique commits adding significant new capabilities:
Run 1Code without any 21st.dev backend connections. Perfect for:
- Privacy-conscious developers who want zero external API calls
- Users who only want to use their own API keys
- Teams with strict network policies
How it works:
- Toggle "Local Mode" in Settings → Preferences
- All cloud features (Automations, Inbox, Remote Sync) are disabled
- Works with Claude, Codex, Custom Models, and Ollama using your own keys
- "Use Locally" quick-start option in onboarding flow
Track your AI spending with detailed analytics:
- Real-time token tracking per model and per chat
- Cost estimation with configurable rates
- Time-range filtering (Today, 7 Days, 30 Days, All Time)
- Per-model breakdown showing input/output/cache tokens
- Activity heatmap for usage patterns
- Secondary metrics: message count, cache read/write, duration
Assign different AI models to different sub-chats within the same project:
- Model inheritance control - choose whether new sub-chats inherit parent's model
- Default model for new chats - set a preferred model globally
- Model persistence - selections survive chat switching
- Custom model profiles - configure multiple providers (OpenRouter, Together AI, etc.)
Configure unlimited custom AI providers:
// Example: OpenRouter profile with multiple models
{
name: "OpenRouter",
baseUrl: "https://openrouter.ai/api/v1",
token: "your-api-key",
models: [
{ name: "Claude 3 Opus", modelId: "anthropic/claude-3-opus" },
{ name: "GPT-4 Turbo", modelId: "openai/gpt-4-turbo" },
{ name: "Gemini Pro", modelId: "google/gemini-pro" }
]
}Features:
- Multiple models per profile
- Offline/Ollama profile pre-configured
- Visibility toggle per model
- Profile validation and deletion
Soft-delete chats without permanent loss:
- Archive popover with file stats (additions/deletions)
- Search and filter archived conversations
- Batch archive multiple chats
- Restore functionality - unarchive when needed
- Archive count indicator in sidebar
Get insights into your conversations:
- ChatDetailsPopover - comprehensive statistics per chat
- Model tags displayed on messages
- Agent model tag component showing which model generated each response
- Token usage per message
- Iterations tracking and context window estimation
Better Windows support for cross-platform users:
- Shell type selection - Choose Bash, PowerShell, or CMD
- Windows build workflow - Automated CI/CD for Windows releases
- Windows path handling - Fixed absolute path issues in file viewer
- VS-style file tree - Proper folder icons and path normalization
Better UX when things go wrong:
- Persistent error banner - Dismissible but remembers state
- Error state management - Tracks errors across all chat transports
- Stream inactivity timeout - Handles proxy disconnects gracefully
- GPT-5.4 Codex as default model (upgraded from previous version)
- Better model tracking in message metadata
- Enhanced model switching UX
This fork maintains all features from the upstream 1Code:
- Claude Code and Codex in one app
- Instant switching between agents
- Custom models and providers (BYOK)
- Cursor-like desktop app
- Diff previews and real-time tool execution
- Built-in Git client with visual staging
- Each chat runs in its own isolated worktree
- Background execution support
- Branch safety - never accidentally commit to main
- Structured plans with markdown preview
- Clarifying questions before execution
- Extended thinking enabled by default
- Full server lifecycle management
- Plugin marketplace
- Rich tool display
- Background agents (cloud sandboxes)
- Live browser previews
- Kanban board
- Chat forking
- Message queue
- Voice input
- Skills & slash commands
- Custom sub-agents
- Memory (CLAUDE.md support)
- Cross platform (macOS, Windows, Linux)
| Feature | Upstream (v0.0.72) | This Fork (v0.0.90) |
|---|---|---|
| Local Mode | No | Yes |
| Token Monitor | No | Yes |
| Per-SubChat Models | Basic | Advanced |
| Archive with Restore | No | Yes |
| Chat Statistics | No | Yes |
| Windows Shell Selection | No | Yes |
| Persistent Errors | Basic | Advanced |
| Custom Model Profiles | Single model | Multiple models per profile |
| Codex Default | Previous | GPT-5.4 |
| Offline Quick Start | No | Yes |
# Prerequisites: Bun, Python 3.11, setuptools
git clone https://github.com/YOUR_FORK/1codepulll.git
cd 1codepulll
bun install
bun run claude:download # Download Claude binary
bun run codex:download # Download Codex binary
bun run build
bun run package:mac # or package:win, package:linuxNote: This fork can be built completely free. Use Local Mode with your own API keys.
bun install
bun run claude:download # First time only
bun run codex:download # First time only
bun run dev- feat(local-mode): Complete offline operation without backend APIs
- feat(monitor): Token usage analytics dashboard
- feat(models): Per-subchat model selection with inheritance control
- feat(archive): Archived chats with restore functionality
- feat(stats): Chat details popover with comprehensive statistics
- feat(windows): Shell type selection (Bash/PowerShell/CMD)
- feat(errors): Persistent error state management
- feat(codex): Upgraded to GPT-5.4 as default
- feat(models): Enhanced custom model profiles (multiple models per profile)
06a742d feat(sidebar): hide cloud features in local mode
4b89b09 feat(settings): add local mode toggle and account connection UI
b2af931 feat(onboarding): add "Use Locally" quick start option
6ad1f53 feat(core): add local mode infrastructure
bafcc96 feat(settings): add Monitor tab for token usage analytics
cbe010f feat(usage): add token usage analytics router
184d581 feat: Enhanced custom model profiles with multiple models
c2f8aeb feat(terminal): add shell type selection
... and 40 more
We welcome contributions that align with the fork's focus:
- Privacy-first features - More offline/local capabilities
- Model flexibility - Support for more AI providers
- Windows/Linux support - Better cross-platform experience
- Developer tools - Analytics, debugging, monitoring
This project is a fork of 1Code by the 21st.dev team.
- Original License: Apache License 2.0
- Original Authors: 21st.dev team
- Original Repository: https://github.com/21st-dev/1code
Please support the original project by subscribing at 1code.dev if you use cloud features.
Apache License 2.0 - see LICENSE for details.
- Original Discord: discord.gg/8ektTZGnj4
- Fork Issues: Open an issue in this repository for fork-specific bugs/features
- Upstream Issues: For core bugs, consider reporting to the original repo
| Resource | Link |
|---|---|
| Original Project | 1code.dev |
| Original Repo | github.com/21st-dev/1code |
| This Fork | Your fork URL |
| Documentation | CLAUDE.md |
| Build Instructions | See Installation section |
Built with love by the community. Forked from 21st.dev's excellent 1Code project.