How to Use HARPA AI for Free

HARPA AI offers a Demo plan that lets you try key features without entering payment details.

The Demo plan has the following limits:

• Up to 10 messages per day using GPT-3.5, ClaudeAI Haiku, or Gemini web session (refreshes daily)
• 50 total AI commands, vision features, agent actions and page-aware messages (this limit does not refresh - once reached, you cannot use commands or page-aware messages)
• 2 web automations
• 2 custom commands

To use HARPA AI for free, you can use the Web Session Connection, limited to chat messages - up to 10 per day.

For additional features or AI Commands and Automations, you can choose between S or X premium plans, detailed on the pricing page:

• S plans, refreshes AI token balance once per billing period - yearly or monthly.
• X plan, lifetime deal includes a preset AI token balance, unlocks all product features permanently, and serves as a BYOK (Bring Your Own Key) plan. You can use your own API keys or web session connections without paying for tokens.

The most budget-friendly monthly S1 plan gives approximately ~30 YouTube summaries or ~50 page-aware queries per day on GPT-4o-mini. Note that GPT-4o or Sonnet uses 10 times more tokens, so the number of queries will be 10 times less.

1 CloudGPT Megatoken (MT) equals to 1 million GPT-4o-mini input tokens, or approximately 750,000 words of English input text. Precise token costs for each LLM can be seen in the HARPA AI connections list. Prices align with those set by LLM service providers - OpenAI or Claude.

On average, 1 MT provides:

• ~ 150 YouTube summaries (15-minute videos) on GPT-4o-mini, or 15 YouTube summaries (15-minute videos) on GPT-4o.
• ~ 50 GPT-4o Vision requests.
• ~ 3,000 basic chat requests using GPT-4o-mini model.
• ~ 300 page-aware chat requests using GPT-4o-mini mode.

# How to Use Free AI Connections in HARPA AI

You can use free Web Session AI connections (ChatGPT, Claude, Gemini) in HARPA AI to avoid additional costs for API or third-party AI platforms. Access to these connections is unlimited on S and X plans.

This is also helpful for HARPA X (BYOK LTD Plan) users. Once you get HARPA X plan, you always have that connection option and don't have to purchase CloudGPT Tokens.

# Free Web Session Connections in HARPA AI

HARPA AI supports connections to multiple LLM providers through your existing user accounts, including ChatGPT, Claude AI, and Google Gemini. You can find detailed setup guides for each provider through the links.

For frequent HARPA AI users, combining CloudGPT connections with other options offers a cost-effective solution. This combination allows you to stay on the most affordable plan while still accessing advanced features like AI Vision and premium models such as ChatGPT 4o, o1, and Claude Sonnet.

The only limitations you'll encounter are the free usage quotas set by each provider for user queries. By switching between different LLMs in Web Session Connections, you can maximize your usage within these free limits without subscribing to premium services like ChatGPT Plus or Claude Pro.

With a variety of connection options, many find that HARPA AI's S1 tier successfully meets their needs while avoiding additional costs. And if you need x3 more tokens, you can upgrade to the S2 plan.

# Free API Connections in HARPA AI

API connections can also be used for free through OpenRouter - setup instructions.

At the time of writing this post, OpenRouter offers several free AI connections, including Google Gemini 2.0 Flash, Llama 3.1 70B, and Mistral 7B Instruct.

You can find the complete list here: https://openrouter.ai/models?fmt=table&order=pricing-low-to-high

To start using OpenRouter, you may need to add a small initial balance ($1-5). You won't be required to spend it and can save it for paid API services whenever you want.

# Free Local LLMs in HARPA AI

You can install LLM models locally on your computer to avoid extra AI costs. Ollama lets you run Llama 3.3, Phi 3, Mistral, Gemma 2, and other models.

Browse all Ollama models at: ollama.com/library

We've written complete installation instructions here.

This is a great tool to reduce AI costs. However, it requires a powerful PC - some models need substantial hardware resources. We recommend checking the technical requirements beforehand, especially for 10B+ models.