| What it does | Fine-tuned small models for English-to-Bash command generation
| Desktop app that compresses Claude Code and ChatGPT Codex inputs locally to cut token costs by up to 50%
|
|---|
| Stage | Not given
| Not given
|
|---|
| Founded | Not given
| Not given
|
|---|
| Based in | Not given
| NL
|
|---|
| Pricing model | Open source
| Freemium
|
|---|
| Pricing | Not given
| Pro: $4/month billed monthly or $3/month billed annually (for Claude Pro or ChatGPT Plus); Max x5: $20/month billed monthly or $15/month billed annually (for Claude Max x5 or ChatGPT Pro $100); Max x20: $40/month billed monthly or $30/month billed annually (for Claude Max x20 or ChatGPT Pro); Team &
|
|---|
| Key features | - 1.5B and 0.6B fine-tuned models
- 401,975 synthetic training examples
- 4-bit quantized models
- CLI with llama.cpp integration
- Command preview mode
- Covers file operations, text processing, Git, archives, networking
- Training data and checkpoints available
- Apache 2.0 and MIT licensed
- ALFA-updated benchmark evaluation
- Supervised fine-tuning with LoRA
| - Local input compression of tool output and boilerplate
- Reversible compression preserving retrieval of original content
- Real-time savings dashboard and activity tracking
- Eight optional add-ons for further optimization
- Support for Claude Code IDE extensions and agent runners
- Privacy-first architecture with all optimization on-device
- Compatibility with Claude Pro, Claude Max, ChatGPT Plus, and ChatGPT Pro
- Minimal latency overhead of approximately 52 milliseconds
- Support for Windows, macOS, and Linux operating systems
- Project learnings saved to agent instruction files
|
|---|
| What makes it different | - 70.7% pass rate on ALFA-updated benchmark vs 63.7% for competing models
- CPU-friendly quantized models
- Specialized for bash command generation from natural language
- Agentically trained with synthetic data generation and evaluation
| - Achieves 25-50% token reduction on typical Claude Code sessions without requiring model subscription upgrades
- All compression runs locally on your machine with no data sent to servers
- Reversible compression preserves output quality and allows retrieval of original content on demand
- Works transparently with existing IDE extensions and agent runners without workflow changes
- Built on open-source Headroom CLI engine available free on GitHub
|
|---|
| Free plan | Not given
| Not given
|
|---|
| Open source | Not given
| Not given
|
|---|
| Platforms | Not given
| Not given
|
|---|
| Public API | Not given
| Not given
|
|---|
| Community votes | 0 | 0 |
|---|
| DR (Domain Rating by Ahrefs) | 36 ● no change since last check | 15 ▲ +1 (up) since last check |
|---|
| Trust Flow (Majestic) | 6 | 6 |
|---|
| Citation Flow (Majestic) | 32 | 31 |
|---|
| Referring domains (Majestic) | 219 | 135 |
|---|
Headroom