Subscription Providers

Overview

Settings → Subscription Providers is where you connect backends you already pay for somewhere else, instead of paying per API call.

The page has two parts:

• Slashspace at the top. The built-in managed provider. It uses your signed-in desktop session and routes chat through the managed backend, so there is nothing to set up. Manage plan opens your plan page on the web.
• Cursor, Claude Code and Codex below it, one row each in an accordion. These are agent runtimes that run on your own computer. The check icon on a row turns green once that provider is connected (an API key is saved for Cursor, the switch is on for Claude Code and Codex).

Unlike cloud providers, models from these runtimes do not just answer. They can read and edit files in a folder you choose.

The three CLI providers require Pro or higher. On the free plan they are blurred behind a lock with an Upgrade button. The Slashspace row stays available on every plan.

---

How this works, in plain English

If you already pay for Claude (claude.ai) or ChatGPT, you can use that same subscription inside Slashspace instead of buying API credits.

Anthropic and OpenAI each publish a small program that runs on your own computer: Claude Code for Claude, Codex for ChatGPT. Those programs sign in with the same account you use in the browser. Slashspace does not call Anthropic or OpenAI directly for these models. It launches the program on your machine, hands it your message, and shows the reply in a chat node.

What that means for you:

• No API key. You sign in once, in the program itself, and never paste a key into Slashspace.
• No Slashspace credits are used for these chats. The usage counts against your Claude or ChatGPT plan instead.
• The work happens locally. Your message goes from Slashspace to the program on your Mac or PC, and that program talks to Anthropic or OpenAI using your login.
• The model can touch files. It reads, and depending on your settings edits, files in the folder you pick as the agent workspace.
• Your tools come along. Whatever MCP servers are active for that chat are passed through to the agent, alongside the built-in canvas tools.

About the terminal

Installing and signing in happens once, in a terminal. After that, everything is inside Slashspace.

• macOS: press Cmd + Space, type Terminal, press Enter.
• Windows: press the Start key, type PowerShell, press Enter.

Then copy a command from below, paste it in, press Enter, and wait for it to finish. Slashspace also shows these same commands, with a copy button and matched to your operating system, inside the CLI Status card whenever it cannot find the program.

---

Claude Code (uses your Claude subscription)

1. Install the Claude Code CLI

macOS and Linux:

curl -fsSL https://claude.ai/install.sh | bash

Windows (PowerShell):

irm https://claude.ai/install.ps1 | iex

Windows (CMD):

curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

If you already have Node.js installed and prefer npm:

npm install -g @anthropic-ai/claude-code

The first three are native installers and do not need Node.js. Use the npm one only as a fallback.

2. Sign in once

claude

The first run walks you through signing in. Pick the option to log in with your Claude account, so the CLI bills your subscription rather than pay-as-you-go API usage. Once it says you are signed in, you can close the terminal.

3. Turn it on in Slashspace

1. If Slashspace was already running while you installed, quit and reopen it. The app reads your system paths at startup, so a freshly installed program is not visible until it restarts.

2. Go to Settings → Subscription Providers, expand Claude Code, and turn on the master switch. Enabling it fetches the model list from the CLI automatically.

3. Check the CLI Status card. A green check with the version number and the file path means you are ready. Use Re-check after installing, updating or moving the CLI.

If the CLI still is not found, the card shows the install steps again, with a copy button on each command.

Options

• Custom CLI path: only needed when the CLI is not auto-detected, for example a non-standard install location. Leave it empty to auto-detect.
• Permission mode: how much the agent is allowed to do.
  • Default: tools that need permission ask for your approval in chat
  • Accept edits: file edits in the workspace are auto-approved, other tools still ask
  • Plan: read-only planning mode, no changes are made
  • Bypass permissions: all tools run without asking (use with care)
• Max agent turns: how many tool-use rounds Claude Code may take per response. Between 1 and 100, default 10.

The Models list has a Refresh Models button. If the CLI cannot be reached, a default model set is loaded instead.

---

Codex (uses your ChatGPT subscription)

1. Install the Codex CLI

macOS and Linux:

curl -fsSL https://chatgpt.com/codex/install.sh | sh

Windows (PowerShell):

powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"

If you already have Node.js installed and prefer npm:

npm install -g @openai/codex

2. Sign in once

codex login

This opens your browser. Sign in with the ChatGPT account that holds your subscription, and pick the sign-in-with-ChatGPT option rather than an API key, so usage counts against your plan. Once it confirms, you can close the terminal.

3. Turn it on in Slashspace

1. Restart Slashspace if it was open during the install.

2. Go to Settings → Subscription Providers, expand Codex, and turn on the master switch. Enabling it fetches the model list from the CLI automatically.

3. Check the CLI Status card, and use Re-check if it does not show a green check.

Options

• Custom CLI path: only needed when the CLI is not auto-detected.
• Approval policy: when Codex stops to ask you in chat.
  • On request: Codex decides which commands need your approval
  • Untrusted: every command asks first
  • On failure: commands run sandboxed, you are asked only when one fails
  • Never: never asks, everything runs inside the sandbox
• Sandbox mode: how much filesystem access the agent has while running.
  • Workspace write: can edit files inside the selected agent workspace
  • Read-only: can read the workspace but not change files
  • Full access: no sandbox, can touch anything (use with care)

The Models list has a Refresh Models button. If the CLI cannot be reached, a default model set is loaded instead.

---

Cursor

Cursor is the odd one out on this page. It connects with an API key and is billed by Cursor, rather than reusing a subscription you sign into locally.

1. Expand Cursor and click Get your API key here. It opens the cursor.com dashboard, where you can create a key.

2. Paste it into the Enter your API key field and save with the check button.

3. Use Refresh Models in the Models list to pull the current model list from Cursor.

---

Agent Workspace

When you chat with a model from one of these runtimes, a folder picker appears in the chat input. The folder you pick is the agent's workspace: the directory it reads from and, depending on permission mode, writes to.

You can also set a canvas-level default workspace in the canvas settings, so every agent chat on that canvas starts in the same folder.

If you never pick a folder, the agent falls back to the folder your canvas file lives in, inside your Slashspace files. That is rarely what you want for real work, so pick the project folder before you start.

---

Caveats

Worth knowing before you rely on this:

• One account per CLI. Whichever account you signed in as in the terminal is the account Slashspace uses. There is no account switcher in the app, so changing accounts means signing out and back in from the terminal.

• An API key in your environment can win over your subscription. Slashspace passes your system environment to the CLI unchanged. If you have something like ANTHROPIC_API_KEY or OPENAI_API_KEY set, or you originally configured the CLI with an API key, it may bill that key instead of your subscription. Run the CLI once in a terminal to confirm which account it is using.

• Restart after installing. Slashspace reads your system paths when it launches. Install the CLI, restart the app, then hit Re-check.

• Unusual installs need a custom path. If you installed through a version manager (nvm, asdf) or into a custom directory, auto-detection can miss it. Paste the full path into Custom CLI path. On Windows, npm installs a small launcher script rather than a real program. Slashspace maps that to the underlying file automatically, but if a launch fails with a launcher-script error, point the custom path at the real executable or reinstall with the native installer.

• File access is real access. Accept edits, Bypass permissions and Codex's Full access sandbox let the agent change files without asking. Start with Plan or Default, keep the workspace pointed at a folder you have backed up or under version control, and widen permissions only once you trust what it is doing.

• The model list can go stale. It is fetched from the CLI, and a default set is seeded when the CLI cannot be reached. Picking a model your plan does not include only fails when you send a message, so use Refresh Models after upgrading or changing plans.

• Agent replies are slower. These runs are multi-step: the agent reads files, calls tools, and may pause for your approval in chat. Max agent turns caps how long that loop can run.

• Install commands come from the vendors and can change. The commands above match what the app shows today. If Anthropic or OpenAI changes theirs, the copy buttons in the CLI Status card are the source of truth.

• Other features still cost credits. Skipping credits applies to these agent chats only. Image generation, RAG indexing and Slashspace-provider chats are billed as usual.

---

Use cases

• Ask questions about a codebase: point the workspace at a repo and use Plan or Default mode to explore and explain code with no risk of changes.
• Delegate real edits: switch to Accept edits and have Claude Code fix a bug or add a feature directly in your project folder, straight from a chat node.
• Get more out of what you already pay for: run Claude Code on your Claude plan and Codex on your ChatGPT plan, with no extra Slashspace credits spent on those runs.
• Keep long tasks bounded: raise Max agent turns for complex multi-file refactors, or keep it low so quick questions never spiral into long tool loops.