Give it the task. It orchestrates the rest.
Ridvay Code is an AI coding orchestrator. Hand it a goal and it plans the work, splits it into subtasks, writes the code, runs the tests, and fixes what breaks — on its own, right in your editor. Powered by Claude or your local models. Zero telemetry, and you can prove it.
Works with Claude · Ollama · LM Studio
Planned 3 subtasks — running them now:
- ✓ Added rateLimiter middleware (token bucket)
- ✓ Wired it into 4 routes
- ✓ Wrote rateLimiter.test.ts · 6 cases
export const rateLimiter = (opts) => bucket(opts.max, opts.windowMs);
Task complete — 6/6 tests passing ✓
Bring your own model
Run Claude for frontier power, or local models for full privacy — your account, your compute. Nothing routes through our servers.
Claude Code
Frontier power, your account
Run the orchestrator on Claude for top-tier reasoning. Bring your own Claude — your account or API key — with no Ridvay middleman in between.
Learn more →Ollama
Local llama & friends
Run Qwen, DeepSeek, Llama and other open models locally with a single command. Fully offline, fully free.
Learn more →LM Studio
Your local model server
Point Ridvay at LM Studio's local server and use any GGUF model on your own hardware, fully offline.
Learn more →Built to finish the job
Most AI tools hand you a suggestion and stop. Ridvay Code keeps going — planning, running, and correcting until the task is done.
Plans, then orchestrates
Decomposes your goal into subtasks, runs each one to completion, and stitches the results back together.
Runs the full loop
Edits, runs commands, reads the output, and adjusts — without stopping to ask at every step.
Fixes what breaks
Catches failing commands and tests and keeps working the problem until it's actually resolved.
Tests until it's green
Writes and runs tests, then iterates on its own code until they pass.
Works across the whole repo
Reads and edits many files at once, keeping a large change coherent from end to end.
You hold the leash
Let it run autonomously, or approve step by step. You decide how much rope it gets.
Truly private by design
No cloud account, no data sharing — privacy isn't a setting, it's the architecture.
No telemetry — and you can prove it
We don't ship a tracker at all. No analytics, no usage data, no phone-home — nothing to opt out of. Grep the published extension bundle and check for yourself.
Your code goes only where you choose
Run fully local with Ollama or LM Studio and nothing leaves your device — or use your own Claude, where requests go straight to Anthropic on your account. Never through a Ridvay server.
No account, no keys to us
There's no Ridvay cloud to sign up for and no API middleman. You bring your own compute or your own Claude, and stay in control.
Open source
Apache-2.0 licensed and fully auditable, built on the open Roo Code agent. Inspect exactly what it does.
Hand off the next task
Install Ridvay Code, point it at Claude or a local model, and give it something to finish.