Integuru is an AI agent that builds permissionless integrations by reverse engineering platforms' internal APIs through browser network traffic analysis.
The tool solves the problem of integrating with platforms that lack official APIs by capturing HTTP requests made during user interactions in a browser, then using an AI agent to understand those requests and generate runnable Python code that reproduces the desired action. The approach works by having users record their browser activity in a HAR file along with cookies and a natural language description of what they want to automate. The agent then identifies the specific network request responsible for the action, traces dependencies between requests to extract dynamic values like account IDs, builds a dependency graph showing which requests must run in what order, and finally generates executable Python code that performs the entire sequence.
This tool suits developers and automation engineers who need to integrate with web platforms that don't provide official APIs or SDKs. It works best for straightforward user actions like downloading documents or retrieving data that can be captured through browser interactions. The project explicitly notes this is version 0, an early public release showing the original approach, with newer versions available elsewhere.
The project maintains a CI workflow using GitHub Actions that runs on each push and pull request to the main branch. Unit tests are available and can be run with pytest. The repository includes a Jupyter Notebook interface alongside command-line usage, offering flexibility in how users interact with the tool. The setup process requires configuring OpenAI API access and managing a Python environment through Poetry, with specific recommendations to use capable models like gpt-4o for graph generation and o1-preview for code generation when available.
