Introduction
Last updated
Tempr ships three products on one account: Chat, an agent toolset inside Visual Studio, VS Code and JetBrains IDEs; CLI, the same agent in your terminal; and Gateway, an OpenAI-compatible API for your own provider keys. A fourth, Tempr Code — the agent in its own desktop app — is in development. All are bring-your-own-key (BYOK) — Tempr never fronts provider tokens or takes a cut of them.
Which one do I want?
If you're not sure where to start, this should make it quick:
| You want to… | Use |
|---|---|
| Chat, complete, and refactor code inside Visual Studio, VS Code, or JetBrains IDEs | Chat |
| Run the same agent from a terminal — one-shot prompts, a REPL, or piped into scripts | CLI |
| Point an existing OpenAI SDK at your own provider keys from a script, server, or app | Gateway |
| Get request logs, analytics, and spend visibility for API traffic outside your IDE | Gateway |
| Give the agent its own desktop app, with several chats running at once on isolated worktrees | Code (in development) |
| Several of the above — one account and login covers them all | Any mix, one login |
They all share one account, one login at the Portal, and one Stripe customer if you're on paid plans for more than one.
How Tempr works
Every request Tempr relays — whether it starts in the IDE chat panel, at the terminal, or hits the Gateway's /v1/chat/completions endpoint — goes out using an API key you supplied for the provider you chose. Tempr never pools keys across customers and never resells tokens at a markup. What you pay for is the product wrapped around your own key: the chat UI and agent tooling for Chat and the CLI, or the virtual keys, logs, and analytics for Gateway.
Chat keeps your conversations on our servers only as the Privacy Policy describes: an agent run for 72 hours after it finishes, and, on Standard and higher plans, your synced chats, unless you turn cloud sync off. Gateway's request/response logging is separate, explicit, and per-account toggleable — see content capture for details.
Use any tool with any model
Most coding tools quietly restrict which models you can reach: Claude Code only talks to Claude, Codex CLI only talks to OpenAI, Copilot only offers its own curated list. Point any of them at Tempr Gateway instead and that restriction disappears — Claude Code can run GPT-5, Codex CLI can run Claude, Copilot can run Gemini, all through the exact same interface each tool already gives you. Tools that are already multi-provider (Cursor, Cline, Continue, and the rest) get the same benefit a different way: one gateway key covers every provider, instead of one key per vendor. See the guides below for the specific tool you use.
Explore the docs
POST /v1/chat/completions.
Guide: Claude Code
Break Claude Code out of Claude-only mode — run GPT-5, Gemini, or anything on your allowlist.
Guide: Codex CLI
OpenAI's own CLI, running models OpenAI doesn't make.
Guide: Cursor
Override Cursor's OpenAI base URL to reach every model on your own allowlist.
Guide: Cline
One provider entry, every model — Claude, GPT, Gemini, and more.
Get help
Can't find what you're looking for? Check the general FAQ or the product FAQs (Chat, CLI, Gateway), or contact us directly.