Pickle Browser
<p> Browser for your agent. Runs local in a window you can see </p> <p> <a href="https://www.producthunt.com/products/pickle-browser?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1221616?app_id=339">Link</a> </p>
Last checked:
What is Pickle Browser?
Pickle Browser is an AI-powered agent browser that runs locally on your machine, letting you watch and control every step of the agent's actions in real-time. It searches pages, reads content, and reports findings while maintaining full privacy and transparency. You can use a bundled free local model, free hosted alternatives, or connect your own Claude, ChatGPT, or MCP-compatible AI.
Pickle Browser pricing
Pricing model: Freemium
Free with bundled local model (no account). Free hosted models via Ollama Cloud or NVIDIA Build (one-time setup). Or bring your own API key for Claude, ChatGPT, or Gemini.
Pickle Browser pros
- Runs locally with full transparency—you watch every step and can take control anytime
- Dramatically token-efficient: uses 156× fewer tokens than raw HTML for page reading (median measured across real sites)
- Flexible AI options: free bundled local model, free hosted models (Ollama Cloud, NVIDIA Build), or bring your own Claude/ChatGPT/Gemini via API
- Privacy-first with local models: page content never leaves your device when using bundled model; no accounts or subscriptions required
Pickle Browser cons
- Local models are unreliable for complex multi-step research—can lose focus and go off-topic; app labels them and warns you
- Windows-only (no mention of macOS or Linux support)
- Free hosted models require sending page text to third-party servers, trading privacy for capability
Frequently asked questions about Pickle Browser
Can I use models I already pay for?
Yes. Connect your existing Claude, ChatGPT, or Gemini API key, or use it as an MCP agent from Claude Desktop, Claude Code, Cursor, or VS Code with 72 available tools.
What happens to my browsing data and logins?
Your pages and history stay local on your machine. With the bundled model, nothing leaves your device. With hosted models, only the page text for that task is sent to the server; your logins and history remain local.
Can I stop the agent mid-task and take over?
Yes. You watch every step in real-time and can take control at any point to manually continue or correct the agent's actions.
How much does it cost to use?
$0 per run with the bundled local model. Using your own paid API (Claude/ChatGPT) costs your normal API rates but uses dramatically fewer tokens than alternatives.