01 · Portal
One portal, every surface
- Keys, usage and agents managed in one place
- The same agent in your editor and your terminal
- Sessions follow you between surfaces
AI coding agent · Now on the VS Code Marketplace
One agent.Any model.Every editor.
Pick Claude, GPT or Gemini per task from one account, and run the same agent inside the editor and terminal you already use.
$ code --install-extension ketecode.kete-code # install$ kete login # sign in
~/acme-api — kete
Works with
- Anthropic
- OpenAI
- Google Gemini
- OpenAI-compatible
Why Kete Code
Many strands, one agent.
Models, surfaces and keys woven into one account, without replacing the tools you already use.
02 · Gateway
Model freedom
- Claude, GPT, Gemini and OpenAI-compatible models
- Bring your own provider keys
- Per-request usage you can actually read
03 · Extensions
Your editor, not a fork
- A native extension, not a forked IDE
- Keep your themes, keybindings and extensions
- Update the agent without migrating your setup
Open source
Open source at the core.
The agent, the kete CLI and the editor extensions are MIT-licensed on GitHub.
Inspect the code
Read exactly what the agent does with your files, commands and context.
Your keys, your models
Run it with your own provider keys or local models through Ollama, LM Studio or vLLM.
No lock-in
The runtime is yours to run, fork and audit, whatever you decide about the hosted services.
Built on OpenCode (MIT). The portal, the Kete Model Gateway and other hosted services are run by Kete Code and are not open source. About OpenCode
Use Kete Code everywhere
Start anywhere, pick up where you left off.
Agent modes
The right mode for the job.
Ask before you touch anything. Plan before you edit. Then let it code, or chase down the bug.
Ask about the code in front of you
Read-only. Kete searches the repo, reads the files that matter and answers with file and line references. Nothing is changed.
kete · panelask
Model gateway
Route every request to the right model.
One account, every provider. Choose a tier as the default and override it per mode or per request.
Balanced
DefaultCost-aware routing for everyday work: reading, small edits, tests.
- Fast, efficient models by default
- Escalate per request when needed
- Usage metered per token
Frontier
Most capableThe most capable models from each provider for large refactors and hard bugs.
- Claude, GPT and Gemini flagships
- Long-context tasks
- Pick per mode or per request
Bring your own
Your keysUse your provider keys or any OpenAI-compatible endpoint through the same agent.
- Anthropic, OpenAI, Gemini keys
- Self-hosted or third-party endpoints
- Usage still visible in the portal
Portal
Keys, usage and models. One place.
Run kete login once. Every surface shares the same keys, limits and usage history.
app.ketecode.ai/usage
Tokens by day
Last 14 days · all surfaces
Balanced Frontier
Default model
- Balancedcost-aware routing
- Frontiermost capable
- Your keyOpenAI-compatible
API keys
| Name | Key | Created | Last used |
|---|---|---|---|
| laptop | kete_sk_…a41f | Sep 2 | 2m ago |
| ci-runner | kete_sk_…09bd | Aug 19 | 1h ago |
| anthropic (own key) | sk-ant-…•••• | Aug 12 | today |
Agents
Soon- reviewer · read, git
- migrations · plan, code
- on-call · debug
Configure once, sync to every surface.
Blog
From the blog
Plan before you edit
Reviewing an agent's plan is cheaper than reviewing its diff. Here's how Plan mode works.
Kete Code Team
One account, many models: how the gateway routes
Why we route per request, and how Balanced, Frontier and your own keys fit together.
Kete Code Team
An AI coding agent that runs inside your editor and terminal. It reads your repo, plans changes, edits files and runs commands, using the model you choose through the Kete gateway.
There will be a free tier. Paid plans add higher limits and team features. See the pricing page for current plans.
Install the extension from the VS Code Marketplace, or from Open VSX in Windsurf, Cursor and VSCodium, and install the CLI with one command. The install page has the steps for each; then sign in once with kete login.
VS Code, plus Windsurf, Cursor and VSCodium through Open VSX, and the terminal. JetBrains IDEs (IntelliJ IDEA, PyCharm, WebStorm, GoLand and more) are coming to the JetBrains Marketplace soon; join the waitlist on the JetBrains page.
The runtime is: the agent, the kete CLI and the VS Code and JetBrains extensions are MIT-licensed on GitHub (github.com/kete-org/ketecode) and built on OpenCode. The portal, the model gateway and other hosted services are proprietary.
Models from Anthropic (Claude), OpenAI (GPT), Google (Gemini) and any OpenAI-compatible provider, including local models served by Ollama, LM Studio or vLLM. The models page lists what is currently available.
Yes. Add provider keys in the portal and Kete routes requests through them. Usage still shows up in one dashboard.
Kete sends the context needed for a request to the model provider you selected. The privacy policy covers what is stored, for how long, and how to have it deleted.
If you sign in to the Kete portal with Google, we receive only your name, email address and profile picture, and use them to create your account and sign you in. We don't access your Gmail, Drive or other Google data, and we never sell it or use it for advertising or model training. See the privacy policy for details.
Ask (read-only questions), Plan (a reviewable plan before edits), Code (edits and commands with approval) and Debug (reproduce and fix). Custom modes are coming soon.
Kete searches and reads files on demand rather than stuffing the whole repo into a prompt, so it works on large codebases while keeping context focused.
The Team plan adds shared billing and centralized key management. Talk to us about enterprise requirements.
Kete runs as a native extension rather than a forked IDE, and lets you switch model providers from one account.
Ready to ship faster?
Install in a minute. Keep your editor. Choose your model.
$ npm i -g @ketecode/cli && kete login