Ridvay Code - Autonomous AI Coding Orchestrator for VS Code

TL;DR

TL;DR: Ridvay Code is a VS Code extension that autonomously plans, writes, tests, and fixes code — running the full development loop without stopping for permission at every step. It works with Claude Code, Ollama, or LM Studio, with zero telemetry by design.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: ridvay.com/ridvay-code ← MUST visit and verify
• VS Code Marketplace: marketplace.visualstudio.com/items?itemName=ridvay.ridvay-code
• Source repository: github.com/tom-tgr/ridvay-code ← verified exists
• License: Apache-2.0 (verified from product page footer)
• HN launch thread: news.ycombinator.com/item?id=43894886 (3pt — very recent launch)

What Is Ridvay Code?

Ridvay Code is an autonomous AI coding orchestrator for VS Code. Give it a goal and it decomposes the work into subtasks, writes the code, runs the tests, and fixes whatever breaks — entirely on its own, right inside your editor.

Unlike AI assistants that hand you a suggestion and stop, Ridvay keeps going. It plans, executes, reads the output, and corrects — without pausing to ask permission at every step.

The product is built as a VS Code extension with a landing page at ridvay.com/ridvay-code. It is open source under the Apache-2.0 license and is built on the open Roo Code agent framework.

Model flexibility

Ridvay Code does not lock you into a single provider. It supports:

• Claude Code — bring your own Anthropic API key for frontier-level reasoning
• Ollama — run Qwen, DeepSeek, Llama, and other open models fully offline
• LM Studio — point Ridvay at a local GGUF model server for total privacy

Requests go directly to your chosen provider. Nothing routes through Ridvay’s servers.

Setup Workflow

Prerequisites

• VS Code (latest recommended)
• One of: Claude Code + Anthropic API key, Ollama running locally, or LM Studio

Step 1: Install the extension

Open VS Code and install from the Marketplace:

# Or search "Ridvay Code" in the VS Code Extensions view

The extension ID is ridvay.ridvay-code.

Step 2: Configure your model

After installation, Ridvay prompts you to select a model provider:

For Claude Code (recommended for best results):

1. Install Claude Code CLI if not already present
2. Ridvay routes directly to it — no additional config needed

For Ollama:

1. Ensure Ollama is running locally (default: http://localhost:11434)
2. Ridvay auto-detects available models

For LM Studio:

1. Start LM Studio and enable the local server
2. Note the server URL and port

Step 3: Give it a task

Open the Ridvay panel in VS Code (Cmd/Ctrl + Shift + P → “Ridvay: New Task”) and describe what you want to build. Ridvay will:

1. Decompose the goal into subtasks
2. Execute them one by one
3. Write and run tests automatically
4. Fix any failures until tests pass

Deeper Analysis

How it differs from existing AI coding tools

Most AI coding tools operate in a stop-and-wait pattern: generate a suggestion, show it to the user, wait for approval. Ridvay inverts this — it runs the full loop autonomously and surfaces results.

This is similar in spirit to what Cursor’s Agent mode or Claude’s Code Agent mode do, but as a VS Code extension with explicit model-agnostic support. The key differentiator is the “you hold the leash” control model — you can switch between fully autonomous execution and step-by-step approval at any time.

Privacy architecture

The product markets itself heavily on privacy, with three concrete claims worth noting:

1. No telemetry at all — the product page states there is no analytics, no usage data, no phone-home. The claim is that grepp-ing the extension bundle would confirm this.
2. No Ridvay cloud — there is no cloud service to sign up for. You bring your own compute or API key.
3. Direct provider requests — when using Claude, requests go straight to Anthropic on your account. When using Ollama or LM Studio, requests never leave the device.

These are architecture claims that align with how the product is designed. Verification requires auditing the extension bundle, which is open source.

Built on Roo Code

Ridvay Code is built on the open Roo Code agent framework (Apache-2.0). This means the underlying agent logic is auditable and not a black box.

Practical Evaluation Checklist

• [ ] Extension installs cleanly in VS Code
• [ ] Claude Code provider configured and working
• [ ] Ollama/LM Studio fallback works offline
• [ ] Task submitted, subtasks planned and executed
• [ ] Autonomous mode vs. step-by-step mode toggled correctly
• [ ] Test writing and auto-fix loop runs end-to-end
• [ ] No requests sent to Ridvay servers during local model use

Security Notes

• No account required — no central database of users
• No telemetry by product design — verify by auditing the extension bundle
• Open source (Apache-2.0) — agent logic auditable on GitHub
• Bring your own API key — credentials never pass through Ridvay infrastructure

FAQ

Q: Does Ridvay Code work fully offline? A: Yes, when configured with Ollama or LM Studio. Nothing leaves your device. With Claude Code, requests go directly to Anthropic on your own API key.

Q: How does it compare to GitHub Copilot? A: Copilot suggests completions inline. Ridvay operates at the task level — it plans multi-step changes, executes them, and iterates until tests pass.

Q: Is the code sent to Ridvay’s servers? A: No. The product has no cloud service. With local models, code never leaves the machine. With Claude, it goes directly to Anthropic.

Q: What license is it under? A: Apache-2.0, and it is built on the open Roo Code agent framework.

Q: Does it require a powerful machine? A: It depends on the model. Local models via Ollama or LM Studio run entirely on your hardware. Claude Code uses remote inference and works on any machine.

Conclusion

Ridvay Code is an autonomous coding agent that runs the full development loop — plan, write, test, fix — directly inside VS Code. Its model-agnostic design (Claude, Ollama, LM Studio), zero-telemetry architecture, and Apache-2.0 open source base make it a genuinely private option for developers who want an AI that finishes the job rather than one that hands off at every step.

If you want an agent that can take a feature request and produce a passing test suite with minimal hand-holding, Worth trying the VS Code extension and seeing how it handles your codebase.