Report Abuse

Basic Information

Dendrite Python SDK is a developer-focused framework for building web-capable AI agents that browse and interact with websites like humans. It provides an asynchronous client API (AsyncDendrite) to navigate pages, click and type into elements, fill fields, press keys, extract structured data, convert pages or fragments to markdown, authenticate on sites and save cookies, and download or upload files. The README includes concrete examples such as automating Outlook to send email, scraping Y Combinator company listings, extracting metrics from Google Analytics, and downloading bank transactions. The SDK supports scaling via a remote browser client that integrates with Browserbase, and it can be combined with external LLMs by configuring API keys such as Anthropic. The project is open source but not under active development; it remains available for forking, study, and adaptation.

Links

Categorization

App Details

Features
The SDK exposes high-level asynchronous primitives for web automation and data extraction, including goto, click, fill, press, wait_for, markdown, extract, get_download and save_as. It supports authentication workflows with a CLI command that opens a browser and persists cookies for later use. The extract facility can produce structured outputs and can cache generated web interaction scripts so they are reused and can self-heal when sites change. Remote execution is supported via AsyncDendriteRemoteBrowser and Browserbase API keys to run many agents in parallel. Examples demonstrate integration patterns for email automation, search navigation, scraping lists into lists of dicts, reading analytics metrics, and downloading transaction files for downstream LLM analysis. Installation is via pip install dendrite and an initial dendrite install command.
Use Cases
Dendrite helps developers give AI systems direct, programmatic access to the live web without depending solely on site APIs. It simplifies common agent tasks such as logging in to services, interacting with complex page elements, extracting clean structured data for use in retrieval-augmented pipelines, and obtaining files for offline analysis. The caching and self-healing of interaction scripts reduce brittle automation maintenance. Remote browser support enables scaling many agent instances without self-managed infrastructure. The SDK provides practical examples and patterns so engineers can quickly prototype agents that combine web actions with LLM reasoning, for tasks like research, reporting, data collection, and automated workflows requiring authenticated access to user services.

Please fill the required fields*