Documentation

Orbit Provider is the first-class way to use managed models without pasting API keys. Sign in with GitHub once, and Orbit loads the models your account can access — including launch promotions like GLM-5.2 at 50% off.

Sign in

• –Open Settings → Providers — Orbit Provider is featured at the top
• –Click Sign in with GitHub and finish auth in your browser
• –You return to the editor; your session stays on your machine
• –Hit Refresh models anytime to pull the latest catalog

Billing & usage

Usage is billed from your prepaid USD wallet on orbiteditorai.com. In the editor you can see remaining credits in Settings → Account and in the chat sidebar profile footer. A low-balance warning appears before you run out mid-session.

Billing & usage opens your web account for top-ups, invoices, and subscription management. Free, Pro, and other plan labels stay consistent across settings and the sidebar.

GLM-5.2 launch offer

GLM-5.2 (Z.ai) is the default Orbit Provider model and is available at promotional rates during the launch window. See Models & pricing for the current per-token rates.

You can still connect Anthropic, OpenAI, OpenRouter, SuperGrok, Ollama, and other providers alongside Orbit Provider and switch per chat.

These are the managed models available through Orbit Provider after you sign in with GitHub. Prices are USD charged to your prepaid wallet per 1 million tokens (input / output). The catalog can grow — use Refresh models in Settings if you don’t see a newly added model.

How to pick a model

• –orbit/GLM-5.2 — default. Strong for agent workflows and long-context coding; currently the launch promo model
• –orbit/deepseek-v4-flash — lowest cost; good for fast Ask / Plan loops and lighter Agent tasks
• –orbit/kimi-k3 — higher-capacity option when you want more headroom on hard multi-file work

Capabilities

Every Orbit Provider model below supports tool calling and reasoning. max_ct means the gateway uses max completion tokens where required by the upstream API.

• –tools — Agent mode can call editor tools (read, edit, search, shell)
• –reasoning — reasoning / thinking modes are supported when the model exposes them
• –max_ct — max completion tokens wiring for OpenAI-compatible upstreams

Billing notes

• –You are only charged for Orbit Provider usage — bring-your-own keys never bill your Orbit wallet
• –Wallet balance is visible in Settings → Account and the chat profile footer
• –Top up from Pricing or Billing on orbiteditorai.com

Prefer your own keys or local inference? Orbit still supports that fully. From Settings → Providers, expand Bring your own provider or Local. Every provider below can be added at the same time; you switch between them per chat.

Cloud providers (API key)

Sign up with the provider, generate an API key from their console, and paste it into Orbit's provider settings.

• –Anthropic — Claude models, API key from the Anthropic console
• –OpenAI — GPT models, API key from the OpenAI platform
• –DeepSeek — API key from the DeepSeek platform
• –OpenRouter — routes to many models through one key (including z-ai/glm-5.2), has its own rate limits
• –Gemini — Google's Gemini models, API key from Google AI Studio
• –Groq — fast inference, API key from the Groq console
• –Grok (xAI) — API key from the xAI console
• –Mistral — API key from the Mistral console, also supports fill-in-middle autocomplete

Sign-in providers (no API key)

• –Orbit Provider — GitHub sign-in, managed models and wallet billing (see above)
• –OpenAI — ChatGPT Plus/Pro — sign in with your ChatGPT subscription to use Codex
• –xAI — SuperGrok / eligible X Premium — sign in to use Composer 2.5 and Grok 4.5 from your subscription

Local providers (self-hosted)

Run models on your own machine or network. No API key — just an endpoint. All four are OpenAI-compatible under the hood.

• –Ollama — default endpoint http://127.0.0.1:11434, supports autocomplete
• –vLLM — default endpoint http://localhost:8000, supports autocomplete
• –LM Studio — default endpoint http://localhost:1234
• –LiteLLM — point at your own proxy endpoint, supports autocomplete

Enterprise cloud providers

• –Google Vertex AI — requires a Google Cloud project, region, and service-account authentication
• –Microsoft Azure OpenAI — requires a project/resource name and API key
• –AWS Bedrock — not called directly; connect through a LiteLLM proxy or the AWS Bedrock-Access-Gateway

Custom endpoints

• –OpenAI-Compatible — any OpenAI API-shaped endpoint, key, and custom headers (use this for llama.cpp and similar)

Example: Ollama

Settings → Providers → Local → Ollama
Endpoint: http://localhost:11434/v1
Model: llama3.2

Inline autocomplete (fill-in-middle) isn't available on every provider — currently Ollama, Mistral, vLLM, LM Studio, OpenRouter, LiteLLM, and OpenAI-Compatible support it. Chat and Agent modes work with all providers above.

Orbit has three chat modes, each suited to a different kind of work. Modes work the same whether you use Orbit Provider or your own keys.

Ask

Get answers about your codebase. Reference files with @, highlight code, or ask general questions grounded in your project.

Plan

For larger changes. Orbit explores your repo, asks clarifying questions, and produces a plan you review before any edits happen.

Agent

Hands-on work across files. The agent can search, edit, and run commands — you approve changes before they land.

Common shortcuts in Orbit Editor.