Connect a model

The assistant runs on the AI model of your choice. The first time you open the panel, it shows Connect a model: pick a provider, paste a key, and you are chatting.

The four providers

Each provider has a step-by-step guide right on the screen, with the exact menus to click as chips (API keys › Create Key) and a link to the right page. You can connect several providers and switch between them from the model picker.

Where the setup lives

• First open: with no key connected yet, the panel opens straight on Connect a model.
• Afterwards: the API keys button (gear) in the panel's header, or the API keys link at the bottom of the model picker. Back to chat returns to your conversation.
• Talios uses the same keys for its coding agent: connect once, use it in both.

Connecting a provider, step by step

Click the provider's tile, follow its guide (shown until the provider is connected), paste the key in the API key field and click Connect. Afterburner checks the key with the provider right away and loads its models; a refused key says so ("Anthropic rejected this key").

Claude (Anthropic)

Gives you Claude Opus, Sonnet and Haiku. Pay-as-you-go, billed by Anthropic.

1. 1

   Sign in to the Claude Console

   Open console.anthropic.com and sign in, or create an account. This is Anthropic's developer console, separate from claude.ai.
2. 2

   Add credits under Billing

   In Billing, add a payment method and buy credits. A Claude.ai subscription (Pro, Max) does not include API usage: the API has its own balance.
3. 3

   API keys › Create Key

   Go to API keys, click Create Key and name it Afterburner.
4. 4

   Copy the key and paste it in Afterburner

   It starts with sk-ant- and is shown only once: copy it before closing the window, paste it in the Claude tile's API key field, click Connect.

ChatGPT (OpenAI)

Gives you the GPT and o-series models. Pay-as-you-go, billed by OpenAI.

1. 1

   Sign in to the OpenAI platform

   Open platform.openai.com, not chatgpt.com: the developer platform has its own account settings and billing.
2. 2

   Add credits under Billing

   In Billing, add a payment method and credits. ChatGPT Plus or Pro does not include API usage.
3. 3

   API keys › Create new secret key

   Go to API keys, click Create new secret key and name it Afterburner.
4. 4

   Copy the key and paste it in Afterburner

   It starts with sk- and is shown only once. The newest models and their visible reasoning can require verifying your organization first: Settings › Organization › General.

Gemini (Google)

Gives you the Gemini Pro and Flash models. Has a free tier.

1. 1

   Sign in to Google AI Studio

   Open aistudio.google.com with your Google account.
2. 2

   Get API key › Create API key

   Go to API keys, click Create API key, then pick or create a Google Cloud project for it.
3. 3

   Copy the key and paste it in Afterburner

   It starts with AIza. The free tier works but has low rate limits, and Google may use free-tier prompts to improve its models: set up billing on the key's project for the paid tier.

OpenRouter

One key for hundreds of models from every lab (Claude, GPT, Gemini, Llama, DeepSeek…), paid from a single balance.

1. 1

   Sign in to OpenRouter

   Open openrouter.ai and sign in with Google, GitHub or email.
2. 2

   Buy Credits

   In Credits, top up your balance. One balance pays for every model OpenRouter serves.
3. 3

   Keys › Create Key

   Go to Keys, click Create Key and name it Afterburner. You can set a spending limit on it.
4. 4

   Copy the key and paste it in Afterburner

   It starts with sk-or-. Only models that can use tools show up in the picker, since the assistant needs them to read your journal.

Connect as many providers as you like: each tile shows a check once connected, and the model picker lists the models of all of them.

Managing a connected key

• Your key follows your account to all your devices, stored encrypted. Signing out removes it from that device.
• You pay your provider directly. Each question goes to your provider with the trades it needs, and the usage is billed on your account there. Afterburner adds nothing on top and does not cap how much you ask.
• Answers keep going without you. Close the app, lock your phone or switch chats: the answer is still written, and waits for you on any device.

The model picker

Bottom left of the message box, the picker shows the provider's logo and the model. Click it to Search models and pick any chat model your keys give access to, grouped by provider and sorted by name, and to set the reasoning effort. Your provider, your model for each provider and the effort are remembered on your account.

Reasoning effort

Higher effort means longer, more thorough answers and a higher bill from your provider. If a question needs steps too long for the server, the assistant lowers the effort on its own before giving up with a message.

What a question costs

• The model you picked answers, billed by your provider like any API call: the more trades it reads and the higher the effort, the more it costs.
• Web searches run on a cheaper model of the same provider when your key offers one (a Sonnet, a GPT mini, a Gemini Flash).
• Free tiers (Gemini) work but hit rate limits quickly on longer analyses.

When something goes wrong