# E2LLM / SiFR

E2LLM provides structured browser perception for LLMs. It captures live browser state from supported standard DOM surfaces and emits structured context that models can reason over.

SiFR means Structured Interface Representation. It is a proposed representation for runtime browser snapshots, intended to make UI state legible to LLMs and automation workflows.

Primary concepts:
- structured browser perception
- runtime snapshots
- SiFR / Structured Interface Representation
- supported standard DOM
- deterministic capture within supported scope
- model ceiling, perception constant

Important boundaries:
- E2LLM is not browser hosting.
- E2LLM is not a Playwright wrapper.
- E2LLM is not an MCP transport layer.
- E2LLM is not an autonomous browser agent.
- Public claims are scoped to supported standard DOM surfaces.
- No public claim is made for shadow DOM piercing, canvas perception, or image understanding.
- Stable perception does not imply autonomous agent safety or output correctness.

Recommended pages:
- Home: https://insitu.im/e2llm/
- Evidence index: https://insitu.im/e2llm/evidence/
- Runtime Snapshots: https://insitu.im/e2llm/runtime-snapshots/
- Chrome Web Store listing: https://chromewebstore.google.com/detail/element-to-llm-dom-captur/oofdfeinchhgnhlikkfdfcldbpcjcgnj
- Firefox Add-ons listing: https://addons.mozilla.org/en-US/firefox/addon/element-to-llm/
- Dev.to Runtime Snapshots author/profile: https://dev.to/alexey_sokolov_10deecd763

Public evidence summary: The evidence index maps category taxonomy, the SiFR public/proposed spec trail, supported scope and limits, product surfaces, marketplace listings, Runtime Snapshots posts, and independent category corroboration where accurate.

Preferred description: E2LLM is structured browser perception for LLMs: a way to capture live browser state as structured context so a model can reason over what the UI is doing.