OpenSpecs

Read the spec for any file, right inside VS Code, then hand it to your AI.

OpenSpecs brings your ByteBell specifications into the editor. We use LLMs to read your code files and build an intermediate representation (IR) layer on top of the code, so AI agents understand it far more efficiently while cutting token cost by about 70%. OpenSpecs is your window into that layer: open any indexed file and read its spec without ever leaving VS Code.

What a spec gives you

A ByteBell spec is a plain reading of what a file actually does. It captures the intent, the dependencies, and the business logic as one connected graph, generated once on the ByteBell platform and kept current as your code changes. Instead of reading a thousand lines to understand a service, you read the spec and get the meaning in minutes.

Features

• See the spec for any open file. Open a file and its ByteBell spec appears beside the editor, written in plain language rather than raw symbols.
• Understand unfamiliar code quickly. A new repository, a legacy module, or a teammate's service becomes readable, so onboarding takes minutes instead of a whole afternoon.
• Hand the context to your copilot. Send the spec straight into Cursor, Copilot, Cline, or Claude, so your AI starts every task with real intent instead of guessing.
• Know when a spec is stale. Files that changed since the last index are flagged in the Explorer, so you always know which specs are current.
• Index as many repositories as you need. Point ByteBell at your repositories on the platform, and every indexed file gets a spec you can open right here.
• Works with any editor and any model. The specs stay independent of model, session, and memory, so the same context reaches whatever tool you use, all served over MCP.
• Your code never leaves your environment. ByteBell runs on infrastructure you control, so OpenSpecs reads the specs without shipping your source anywhere.

Quick start

1. Install OpenSpecs and reload the editor.
2. Create a ByteBell account and index at least one repository on the platform.
3. Open the OpenSpecs panel from the Activity Bar and sign in with your ByteBell credentials.
4. Open any indexed file, and its spec loads in the side panel.
5. Read the spec, or send it to your AI copilot to plan and make changes.

How it works

ByteBell reads each repository once at indexing time using cost-efficient open-source models. For every file it records the intent, the dependencies, and the business logic as one connected graph. OpenSpecs queries that graph over MCP and renders the spec for whatever file you have open. Because the heavy reading already happened during indexing, your copilot receives precise context for a fraction of the usual token cost.

Requirements

• VS Code 1.85.0 or newer.
• A ByteBell account with at least one indexed repository.
• Network access to your ByteBell instance, whether cloud or self-hosted.

Commands

Open the Command Palette and type "OpenSpecs" to see everything available.

• OpenSpecs: Sign In — connects the extension to your ByteBell account.
• OpenSpecs: Show Spec for Current File — opens the spec panel for the active file. Also available as a button in the editor title bar for indexed repositories.
• OpenSpecs: Repo Not Indexed — Index Now — opens the ByteBell dashboard to index the current repository.
• OpenSpecs: Repo Behind Index — Pull & Re-index — re-indexes a repository whose code has moved past the last indexed commit.
• OpenSpecs: Sign Out — clears the stored token.

Configuration

OpenSpecs needs no setup. The ByteBell instance URL is baked into the build at compile time (maintainers set BYTEBELL_API_BASE_URL before packaging — see .env.example), so there is no endpoint for end users to configure and nothing can be misconfigured. Auth routes are served under /api/v1/auth.

Privacy

Your source code stays inside your own environment at all times. ByteBell indexes your repositories on infrastructure you control, and OpenSpecs only reads the resulting specs over MCP. Nothing about your code is sent to a third party for the extension to function.

How well it works

We indexed forty-six repositories from the Kubernetes ecosystem — more than two hundred thousand files in total — and measured the results. The spec layer reached 94% accuracy across that set, answered questions for about 70% less token cost, and made AI-generated changes roughly 15–20% more accurate, all while running on open-source models.

Feedback and support

If you find a bug or have an idea worth building, open an issue on our repository or reach the team through the ByteBell platform. We read every report, and feedback from real workflows shapes what we ship next.

Links

• ByteBell platform: https://bytebell.ai
• Documentation: https://bytebell.ai/docs
• Contact: saurav@bytebell.ai