Skip to content

About

The VS Code extension for BharatCode (bharatcode.ai, an LLM inference provider): install, sign in, and the bundled agent is ready to run - provider, endpoint and models preconfigured. Fork of Kilo Code.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

77 stars

Watchers

0 watching

Forks

Latest commit

 

History

112 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

BharatCode for VS Code — the VS Code extension for bharatcode.ai

Marketplace version Latest release CI status MIT licensed Provider: bharatcode.ai


Your coding agent, ready in one click

BharatCode for VS Code installs the whole agent for you — runtime, provider, endpoint and models included. Sign in, and you are writing with it.


Install from the VS Code Marketplace »

Or press Ctrl+P / ⌘P and paste:

ext install thenabbu.bharatcode-vscode

VS Code 1.105.1 or newer. Chromebook, Linux, macOS and Windows all get their own build.


Three steps to your first task

1. Click install. The Marketplace link above does everything — VS Code downloads it and adds the agent to your editor. Around 160–180 MB, because the entire runtime travels inside the package. Nothing to fetch afterwards.

2. Open a folder. File → Open Folder. Your project is what the agent reads and edits.

3. Sign in. BharatCode greets you on first launch and offers both ways in:

Sign in with BharatCode API key
Best for Nearly everyone Scripts, shared machines, CI
How A browser tab opens; click approve; done Paste a key from your bharatcode.ai account
Typing None One paste

Both are first-class paths and both work on first run — pick either. Your credential is stored by the extension backend, never in settings.json, so it stays out of your repo and your dotfiles.

Missed the prompt? BharatCode: Connect Account in the Command Palette brings it back.

That is the entire setup. No base URL, no provider JSON, no model list, no CLI.


What it does

Writes across your project. Describe the change; it edits the files, shows you a diff, and you approve what lands.

Switches hats with you. Code implements, Plan designs before writing, Ask answers from your codebase, Debug traces a failure to its cause. Custom agents are plain-text definitions you can write yourself.

Runs several tasks at once. Agent Manager puts sessions side by side, each in its own git worktree, so parallel work stays out of its own way.

Runs what it wrote. It has the terminal and a browser, so it can execute the build, read the error, and fix it before handing back.

Checks its own work. It re-reads what it produced and corrects before you see it.

Learns your project. MCP servers, skills and rules live in plain text and can be scoped per project.

Handles the small stuff. Commit messages written from your staged diff, plus inline actions on any selection: explain, fix, improve, add to context.


The models

Your BharatCode account comes with two models, and the extension wires both up for you. Switch mid-task from the picker in the chat header.

Model Context Best at
qwen-3.8-27b 262k Everyday edits and fast iteration
deepseek-v4.1-flash 1M Whole-repo reasoning and long sessions

Both come from one OpenAI-compatible endpoint at https://bharatcode.ai/api/model/v1, and the list is read live from it.

Prefer another model? Add it yourself. Settings → Providers → Custom provider takes a base URL, a key and a model list, and supports OpenAI, OpenAI-compatible and Anthropic endpoints. Bring your own key for OpenAI, Anthropic, or anything else you run, and it sits alongside the BharatCode models you already have.


Install by hand

Download a .vsix for your platform

Grab the file for your system from Releases, then run Extensions: Install from VSIX... from the Command Palette:

code --install-extension bharatcode-vscode-linux-x64-<version>.vsix

Eight platform builds ship per release, plus SHA256SUMS:

Platform File suffix Platform File suffix
Linux x64 linux-x64 Linux arm64 linux-arm64
Alpine x64 alpine-x64 Alpine arm64 alpine-arm64
macOS Intel darwin-x64 macOS Apple Silicon darwin-arm64
Windows x64 win32-x64 Windows arm64 win32-arm64

Each carries its own runtime binary, which is why they are split per platform — you download yours, once. Verify with sha256sum --check SHA256SUMS.

macOS first run: builds are unsigned for now. If Gatekeeper blocks it: xattr -dr com.apple.quarantine ~/.vscode/extensions/bharatcode.bharatcode-vscode-*


FAQ

Is BharatCode the same thing as this extension?

Two separate things, and the split is what makes setup simple.

BharatCode (bharatcode.ai) is an LLM inference provider — one OpenAI-compatible endpoint, one key, two chat models. It serves models. It is your account.

This extension is the VS Code client for that provider. It ships the agent runtime inside the package and points it at your BharatCode account, which is why signing in is the only step you take.

Install the extension, sign in, and the agent is running — because the extension does the wiring.

Where does this extension come from?

It is a fork of Kilo Code, itself a fork of OpenCode. Those projects and their contributors built the agent this package runs, and the credit is theirs — MIT licensed, with attribution preserved in LICENSE and in the kilocode_change markers throughout the source. This fork brings that agent to the BharatCode provider, rebrands it, and reconfigures the provider layer so bharatcode.ai is wired in from the start. We build on their work; we do not claim it.

Do I need a key if I sign in with my browser?

One or the other is enough. Browser sign-in stores a session; a pasted key stores a key. Both land in the extension backend's auth store and both go only to the provider you connected.

Why two models instead of a big catalogue?

Because that is what your BharatCode account serves, and they cover a lot of ground between them: 262k context for fast day-to-day work, 1M for whole-repo reasoning. If you want more, add your own provider under Custom provider and your own models show up next to these.

How do I sign out, or switch accounts?

Settings → Providers → the BharatCode card → disconnect. That clears the stored credential, so a stale login cannot quietly reconnect. Uninstalling keeps the credential files, so a reinstall resumes your session — for a clean slate, quit VS Code fully and delete ~/.local/share/kilo/ (%USERPROFILE%\.local\share\kilo\ on Windows).

Does it work offline?

The agent runtime lives on your machine, so it starts without a network. The models are served remotely, so plan on a connection for actual work.


Where to look next

If you want to… Read
Change a setting or pick a model docs/CONFIGURATION.md
Know exactly what leaves your machine PRIVACY.md
Contribute code or docs CONTRIBUTING.md
Work on the codebase AGENTS.md
Cut a release RELEASING.md
Report a vulnerability SECURITY.md

Contributing

Pull requests are welcome from developers and writers alike. Start with CONTRIBUTING.md; if you are touching code, read AGENTS.md — it is the same file the agent itself follows. The Code of Conduct applies to every space around this project.

License

MIT — use it, modify it, ship it commercially, keeping the attribution and licence notices. Full text, including upstream copyright lines, in LICENSE.

About

The VS Code extension for BharatCode (bharatcode.ai, an LLM inference provider): install, sign in, and the bundled agent is ready to run - provider, endpoint and models preconfigured. Fork of Kilo Code.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

77 stars

Watchers

0 watching

Forks

Releases

Contributors

Languages