E2LLM / SiFR Evidence Index

This page collects public evidence for E2LLM and SiFR. Its purpose is narrow: help a technical analyst recover the category and the moat in 5–10 minutes without mis-bucketing E2LLM as browser hosting, a Playwright wrapper, an MCP transport layer, or an autonomous browser agent.

Category statement

E2LLM is structured browser perception for LLMs. It captures supported live browser state and emits a model-readable representation. SiFR is the proposed representation layer for those runtime snapshots.

Analyst path

Start with the category taxonomy. Runtime Snapshots #16 explains the difference between vision-based browser agents, accessibility/tree abstractions, and structured runtime perception.

Read the SiFR representation trail. Runtime Snapshots #7 introduces SiFR as the proposed structure that makes browser UI state legible to models.

Check supported scope and limits. Standard DOM surfaces only; no shadow DOM piercing; no canvas/image perception claim.

Check product surfaces. E2LLM site and product docs, MCP tools where present, browser extension listings, GitHub/dev documentation.

Check the repeatability trail. Existing public Runtime Snapshots #10 and #14 show production-loop and pipeline usage. The canonical repeatability demo is owned by the E5 workstream and will be pinned here once staged. repeatability demo: pending E5

Check marketplace / listing proof. Chrome Web Store, Firefox Add-ons, GitHub, Dev.to / Runtime Snapshots.

Check independent category corroboration. ViewGraph / @viewgraph/core references Element to LLM / E2LLM as inspiration / category input for structured perception. Treat as category corroboration, not adoption proof.

Proof map

Category taxonomy

Runtime Snapshots #16 — “The Three Architectures of Browser Agents.” Runtime Snapshots series

SiFR public / proposed spec

Runtime Snapshots #7 — “Inside SiFR: The Schema That Makes LLMs See Web UIs.” Runtime Snapshots series

Product / tool surface

The E2LLM site and product docs, including MCP tools where present. E2LLM home.

Supported scope / limits

Standard DOM surfaces only. No shadow DOM, canvas, or image perception claims.

Repeatability trail

Runtime Snapshots #10 — “The Loop: E2LLM in Production”; #14 — “From Clipboard to Pipeline.” Runtime Snapshots series. Future canonical /repeatability-demo/ once the E5 same-state → same-capture → diff-model artifact ships. pinned slot: pending E5

Distribution proof

Chrome Web Store listing • Firefox Add-ons listing • GitHub • Dev.to Runtime Snapshots series.

Third-party corroboration

@viewgraph/core / ViewGraph reference to Element to LLM / E2LLM, where accurate. Category corroboration, not adoption proof. reference link: pending confirmation

How we describe it (and what we do not claim)

Say	“SiFR is a proposed Structured Interface Representation.”
Say	“Deterministic capture within supported standard-DOM scope.”
Say	“Model ceiling, perception constant.”
Don't	“SiFR is the standard” — not unless and until adoption exists.
Don't	“100% perception”, or imply perception quality from a task-success percentage.
Don't	Claim shadow DOM, canvas, or image support.
Don't	Imply autonomous agents are safe because the perception layer is stable.

For implementation details, read Runtime Snapshots and the product docs. For enterprise evaluation, review this public proof map first, then request a technical walkthrough.