Write, edit, debug, and ship code with AI — powered by local models (Ollama) or cloud APIs. Zero telemetry. Your code stays yours.
Auto-detects GPU, RAM, and model size. Adjusts tool complexity to match your hardware — from MINIMAL (3 tools) to FULL (all tools).
Cloud API hits rate limit? Auto-switches to local Ollama. Re-adapts tool descriptions for the new model. Zero downtime.
Zero telemetry by default. All LangSmith tracing blocked at code level. Your data never leaves your machine.
Real-time token usage and cost per provider. Free with Ollama, from $0.14/M tokens with cloud APIs.
Blocks dangerous operations: rm -rf, git push --force, DROP TABLE. Configurable patterns and sensitive paths.
Polls Ollama every 30s. New model pulled? Auto-discovered. Model removed? Auto-disposed. All in real-time.
/model, /adaptive, /fallback, /registry, /privacy, /providers, /doctor, /benchmark, and 45 more.
100% coverage on all custom middleware. 29 eval scenarios across 4 categories. Battle-tested.
Ollama, Groq, DeepInfra, Together, OpenRouter, Anthropic, OpenAI — switch with one flag.
Roncon Code charges for the CLI tool — you bring your own LLM (local or cloud API).
Proof-of-work throttled
~0.1s first 10 actions/hr
~1s up to 30/hr
~16s up to 50/hr
You bring your own model — local (free) or cloud API (your key, your cost).
| Model Setup | LLM Cost | Quality | Privacy | Speed |
|---|---|---|---|---|
| Ollama local (7B) | $0 (free forever) | Good | 100% local | ~5 tok/s |
| Ollama local (14B+) | $0 (free forever) | Very Good | 100% local | ~3 tok/s |
| Groq (70B cloud) | ~$37/mo | Very Good | API only | ~300 tok/s |
| DeepInfra (72B cloud) | ~$22/mo | Very Good | API only | ~50 tok/s |
| OpenRouter (DeepSeek) | ~$10/mo | Good | API only | ~40 tok/s |
Compare: Claude Code costs $100-200/mo (CLI + model bundled). GitHub Copilot $19/mo.
With Roncon Code, you pay only for the CLI + choose your own model at your own cost.