AI for builders
Your Unix terminal speaks English now.
echo "what you want" | onnai
No Python. No man pages. Just | and go.
Install in 10 seconds
Click to copy • Paste into Terminal
Trial: $10 to try. Plans from $9/mo.
Quick Start
# Charge your Onnai with $10 of credits
onnai --dropin you@email.com
# Use onnai
onnai "explain this error message"
# Pipe anything
git diff | onnai "explain these changes"
ps aux | onnai "explain like I am ten"
ls -al | onnai "roast my directory"
# Add specialized Mimics
onnai --add Coder
cat buggy.py | Coder "find the bug"
# Chat mode - try a Mimic
onnai -m Volcanara
> hi
# Out of credits? --addfunds or subscribe
onnai --addfunds
Why Onnai?
The name is the contract. You call Coder. We handle which model, which prompt, which version. Next month we upgrade what's behind it. Your scripts get smarter. You change nothing.
Compose like Unix. Chain Mimics together:
curl -sL news.ycombinator.com | onnai "top 5 stories" | onnai "explain like I'm 10"
Your files, your machine. Context lives in .context. Conversations in .chat. Edit them, search them, version them. No cloud lock-in.
Bring your own keys. Set OPENAI_API_KEY or ANTHROPIC_API_KEY and go. Same interface, your infrastructure. Or run local with onnai -p mlx.
Mimics
No prompt engineering. The name is the interface.
Coder
Reviews, refactors, fixes bugs, writes tests.
onnai --add Coder
Sys
Unix commands, scripts, system optimization.
onnai --add Sys
Writer
Docs, emails, READMEs. Matches your voice.
onnai --add Writer
Reviewer
Code review. Catches what you missed.
onnai --add Reviewer
Or make your own: drop a .persona file in any directory.
Examples
Code
# Explain unfamiliar code
cat legacy.py | onnai "what does this do"
# Generate tests
cat api.py | Coder "write pytest tests" > test_api.py
# Fix errors
python script.py 2>&1 | onnai "fix this"
Git
# Explain changes
git diff | onnai "summarize"
# Write commit message
git diff --cached | onnai "commit message"
# Code review
git show HEAD | Reviewer
System
# Diagnose issues
ps aux | onnai "anything suspicious?"
# Explain logs
tail -100 /var/log/syslog | onnai "any errors?"
# Get the command you need
onnai "find files larger than 1GB modified this week"
Clipboard workflow
# Copy → transform → paste
pbpaste | onnai "fix grammar" | pbcopy
pbpaste | onnai "make concise" | pbcopy
pbpaste | onnai "to bullet points" | pbcopy
For Developers
Git hooks
# .git/hooks/pre-commit
git diff --cached | Reviewer || exit 1
Log monitoring
tail -f app.log | grep ERROR | onnai "diagnose"
Documentation
find . -name "*.py" | xargs cat | onnai "generate API docs" > API.md
Project context
# .context file in project root
echo "Python 3.11, FastAPI, PostgreSQL. Follow PEP 8." > .context
# AI now knows your stack
cat endpoint.py | Coder "add error handling"
Compare providers
# Same prompt, different models — no accounts needed
onnai -m openai/gpt-5.1 "explain quicksort" | onnai "grade 1-10, explain your reasoning"
onnai -m anthropic/claude-opus-4-5-2025110 "explain quicksort" | onnai "grade 1-10, explain your reasoning"
onnai -m cerebras/zai-glm-4.6 "explain quicksort" | onnai "grade 1-10, explain your reasoning"
onnai -p ollama -m minimax-m2.1:cloud "explain quicksort" | onnai "grade 1-10, explain your reasoning"
Technical
Dual-process design: onnai CLI + onnaid daemon.
onnai (CLI)
Resolves symlinks via argv[0]
coder → onnai
Walks tree for .context / .persona
Unix socket to daemon
onnaid (Daemon)
Event-driven C, non-blocking
kqueue / epoll
Persistent HTTP/2 pool
Zero cold starts
BYOK Providers
Set your API key and go. Traffic never touches onnai.ai.
Cloud
OpenAI · Anthropic · Groq · Cerebras · Together · Mistral · X.AI
Local
MLX · Ollama
No keys required
# Use any provider
onnai -p anthropic "your prompt"
onnai -p mlx -m llama3 "your prompt"
# List available models
onnai -p cerebras -l
Plans
Use with BYOK. Subscribe for managed AI. Use any provider with unified billing.
Creator
- Pick any Mimic
- Create your own Mimic
- Top off when out of Mimic credits
- Hobbyist/Enthusiast Mimic Usage Included
- BYOK Provider Abstraction enables model swap
- BYOK Inference never touches onnai.ai
- Community support
Pro
- Everything in Creator
- Instant Mimic inference (up to 3000 T/sec)
- Unified billing: add models from OpenAI, Anthropic, X.AI, Cerebras, Grok, and others
- Unified:
onnai -m cerebras/zai-glm-4.6 - Private:
onnai -p mlxon your laptop
Max
- Everything in Pro
- Always-on Onnai Machine with SSH access
- Server-based
onnai -p mlx - Scale Compute/Storage/Inference (Pay As You Go) for your API
- Priority support
Custom
- Usage analytics, admin, logging
- On-Premise / Private
- Forward-deployed engineers
- Seminars & Training
All plans include unlimited BYOK usage. Cancel anytime.
Need More?
Private cloud deployment, custom model fine-tuning, or dedicated support? Let's talk.