A developer named Meet Pandya has published Plaidify, an open-source, self-hosted REST gateway designed to give AI agents programmatic access to login-protected websites that lack public APIs. The project, released under an MIT license on GitHub, accepts credentials and a JSON "blueprint" file describing a site's authentication and data extraction flow, then returns structured JSON — targeting use cases like bank balances, utility bills, insurance portals, and government sites. Pandya positions it as a free, self-hosted alternative to Plaid, which covers only financial institutions and starts at over $500 per month. The self-hosted architecture is a deliberate choice: credentials and extracted data never leave the deployer's own infrastructure.
The core technical approach relies on <a href="/news/2026-03-14-verge-browser-self-hosted-isolated-browser-sandbox-for-ai-agents">Playwright for browser automation</a> — the library handles authentication and structured scraping on the operator's behalf. However, the project is at an early, aspirational stage: as of publication, the Playwright engine is a stub returning simulated responses, and real browser integration is flagged as the top-priority open contribution needed. The blueprint system is the more complete part of the stack, allowing operators to drop a small JSON file into a connectors directory to define login fields and data extraction selectors, with Python connectors available for sites requiring custom logic. A three-phase roadmap outlines the REST API and blueprint system in Phase 1, a community blueprint library in Phase 2, and an MCP server with agent SDK and user consent model targeted for Q4 2026 in Phase 3 — the last of which would make Plaidify a native data source for Claude, GPT-based agents, and any MCP-compatible client.
Van Buren v. United States (2021) is the controlling U.S. framework. In that case, the Supreme Court adopted a narrow "gates-up-or-down" interpretation of the Computer Fraud and Abuse Act: a user who authenticates with their own legitimate credentials does not exceed authorized access under the CFAA. That provides some protection for Plaidify's core model. But Power Ventures v. Facebook (9th Cir. 2016) established that a platform's cease-and-desist letter converts previously authorized access into a CFAA violation, regardless of underlying user consent — a precedent that maps closely onto Plaidify's threat model for any operator whose blueprint targets a site that objects. Outside the U.S., the UK's Computer Misuse Act 1990 provides no civil loss-floor equivalent to the CFAA's $5,000 threshold, meaning UK-hosted deployments face categorical criminal exposure the moment a blueprint defeats an authentication gate without the website operator's consent.
No well-maintained open-source library currently handles authenticated web scraping at the level agentic workflows require — structured login, session management, and structured data extraction in a single deployable layer. That's the gap Plaidify is trying to fill, and it's a real one: personal finance agents, life-admin automation, and productivity tools all hit the same wall when structured data sits behind a login form with no API. Whether the project matures into production-ready tooling depends on two milestones — community completion of the real Playwright integration, and the Phase 3 MCP server — both of which remain pending. Developers evaluating Plaidify for <a href="/news/2026-03-15-axe-go-binary-toml-llm-agents-unix-pipes">agentic workflows</a> should treat it as early-stage infrastructure with meaningful legal considerations attached, particularly for any deployment targeting financial institutions or sites that have issued scraping prohibitions.