Cloud Providers

Overview

Settings → Cloud Providers is where you connect AI providers that run in the cloud. The page shows an accordion with one row per provider:

• Rabbitholes — Slashspace's built-in managed provider (no API key needed)
• OpenAI, Anthropic, Google, Perplexity, xAI, OpenRouter, Straico, Qolaba, DeepSeek, Groq, and Fal AI (BETA) — bring your own API key

The key icon on each row turns green once an API key is saved for that provider, so you can see at a glance which providers are connected.

Local models (Ollama), OpenAI-compatible endpoints (Custom Providers), and coding-agent CLIs (Agent Providers) each have their own settings page.

On the free (Starter) plan, only Rabbitholes is available here. The other providers appear blurred behind a lock — "Cloud providers require Pro or higher. Upgrade to use your own API keys with OpenAI, Anthropic, Google, and more." — with an Upgrade button.

---

Rabbitholes (built-in)

Rabbitholes needs no setup at all. As its card explains: "Rabbitholes uses your signed-in desktop session and routes chat through the managed backend endpoint."

• There's no API key to paste — usage is billed against your credit balance.
• The Manage plan link on the card opens your plan page on the web, where you can check or reload credits.

Custom models are not available for the Rabbitholes provider — you can only add your own models to BYOK providers.

---

Adding an API Key

1. Go to Settings → Cloud Providers and expand the provider you want to connect.

2. Click Get your API key here — it opens that provider's dashboard in your browser, where you can create a key.

3. Paste the key into the Enter your API key field. It's a password field; use the eye button to show or hide the key.

4. A footer appears with a Skip API key validation checkbox, a red X (cancel your changes), and a check button (save).

5. Click the check button to save. Slashspace validates the key with the provider before saving, unless you ticked Skip API key validation.

When you save a key for the first time, a default set of models for that provider is added automatically, and if you haven't set a default chat model yet, one is picked for you (you can change it in Default Models).

---

Managing Models

Each provider row has a Models list:

• Toggle — the switch on each model row activates or deactivates the model. Only active models show up in model pickers across the app.
• Delete — the trash button removes a model after a confirmation: "This will remove the model from the active list."
• Add New Model — opens a dialog to register any model the provider supports. For providers whose model inventory can be fetched, a combobox lets you pick from the live list; otherwise fill in:
  • Model ID (required) — the provider's exact model identifier, e.g. gpt-4o
  • Display Name / Alias (optional) — the name shown in the UI
  • Description (optional)
  • Modalities — checkboxes such as Text to Text, Text to Image, Image to Text; the model type is derived from what you select
• Refresh Models — shown instead of the add button for providers that fetch their model list (like agent providers).

Chat models are validated against the provider before they're added, so a typo in the model ID fails fast.

Per-model settings

Click a model card to open its Model Settings dialog. You can set an alias, a model-specific system prompt, and advanced sampling parameters: Temperature, Top K, Top P, Presence Penalty, Frequency Penalty, Max Tokens, and Max Steps.

Provider-level advanced settings

Each provider also has a Manage Advanced Settings button. This dialog sets defaults for the whole provider — System Prompt plus the same sampling parameters and Max Steps / Max Tokens. Every field has a Clear control, and the dialog has Reset Defaults and Save buttons.

---

Use cases

• Start with zero setup: chat with Rabbitholes models right after signing in — no keys, no configuration — and reload credits from the Manage plan link when you run low.
• Bring your own keys to control cost: on Pro, add your OpenAI and Anthropic keys so heavy usage bills your own accounts at provider rates instead of consuming credits.
• Pin a new model the day it ships: when a provider releases a model that isn't in the default list yet, use Add New Model with its model ID and start using it immediately.
• Tune a provider for one workflow: set a provider-level system prompt and a low temperature via Manage Advanced Settings so every model from that provider answers in a consistent, deterministic style.