Use your own API keys in HARPA
Bring your own key: OpenAI, Anthropic, Gemini, OpenRouter and more.
On this page
Add an API key from your AI provider and HARPA chats with its models directly. You pay the provider for what you use, it never logs you out, and it never uses cloud credits. Any OpenAI-compatible endpoint works too, including models you run on your own computer.
# Add an API key
- Get a key from your provider's dashboard: OpenAI, Anthropic, Google AI Studio and so on.
- Open HARPA on any page (Alt+A on Windows, Ctrl+A on Mac), click the model button in the composer and choose All connections. Alt+M on Windows or Ctrl+M on Mac opens the same Models screen.
- Scroll to Your API keys and click Add API key. A new key card opens.
- Pick the provider. HARPA fills in its address and a default model.
- Paste the key into API key. Get API key next to the field opens the provider's key page.
- Check the chat model: type the model's id as the provider names it. Available models opens the provider's list.
- Changes save as you type. Click the sliders button next to Your API keys to see the rows again, and click your key to use it in this chat.
| Provider | Key page | Note |
|---|---|---|
| OpenAI | platform.openai.com | |
| Anthropic | platform.claude.com | |
| Google Gemini | aistudio.google.com | |
| DeepSeek | platform.deepseek.com | |
| xAI Grok | console.x.ai | |
| Perplexity | perplexity.ai, API keys | Answers with live web search |
| OpenRouter | openrouter.ai | One key for models from many vendors |
| Straico | platform.straico.com | Answers arrive whole, not word by word |
| Custom | Your provider or your server | Any OpenAI-compatible endpoint, local models |
# Use a key in your chats
Your keys show under Your API keys with their model and address, like gpt-5.4 · api.openai.com. Click a row to use
it in this chat; new chats start on it too. Keys also appear in the model menu of the composer, next to the cloud
models.
To edit a key later, click the sliders button (Advanced settings) next to Your API keys: the rows turn into cards. Click a card's title to rename it; hover it to clone it, say to use one key with two models, or to remove it.
# Connect a custom endpoint or a local model
Pick Custom as the provider for any service that speaks the OpenAI Chat Completions API, or for a model running on your own computer with Ollama:
- Open Advanced settings inside the key card.
- Set API proxy URL to the server's address without the path, like
http://localhost:11434for Ollama. - Set API proxy path, usually
/v1/chat/completions. - Type the chat model, like
llama3.2. Leave API key empty if the server doesn't need one.
Ollama only answers the extension when it runs with OLLAMA_ORIGINS=chrome-extension://*. The full walkthrough:
Connecting to local LLMs.
Advanced settings also hold headers, a system prompt, the context window, max response tokens, temperature, Stream responses and MCP calls, which lets this key use your MCP servers.
# Good to know
- No credits. Chats on your own key are billed by your provider, never in cloud credits, on every HARPA plan.
- Straight to the provider. Prompts on your own key go from your browser to the provider.
- Saved to your HARPA space. Your keys sync to your HARPA space, so they follow you to your other devices. In a team space every member can chat on them, and only the space owner can add or change them.
- It doesn't answer? Check the key, that billing is on in your provider account, and for Custom, that the URL has no path in it. The error in the chat links to the provider's status page when there is one.
- Rather not manage keys? HARPA's cloud models give you GPT, Claude, Gemini and more without any.
More walkthroughs: OpenAI API key, OpenRouter, DeepSeek.