# E2LLM (llms.txt)

What E2LLM is: E2LLM connects the AI you already use — ChatGPT, Claude, Grok, Perplexity — to your own browser. Your AI reads live pages as structured SiFR and acts on them (click, fill, scroll) in your real, signed-in session, with your approval, over an open protocol (MCP relay, live at mcp.e2llm.com).

Category: structured browser perception + action / runtime perception.
Vendor: insitu (S2 Tikshuv Ltd), Israel.
Site: https://e2llm.com

## Canonical terminology

- SiFR = structured capture format of a rendered page (graph of nodes, selectors, bounding boxes, relationships), see /sifr.
- Runtime perception = capturing the page as it exists in a live session, after JS, with authentication, as opposed to static HTML fetch.
- Runtime Snapshots = the E2LLM article series, see /snapshots.
- Relay = the MCP server (mcp.e2llm.com) that connects an AI client to the user's own browser for perception and action.

## Canonical links

- /developers — developer hub: how perception works, determinism, client compatibility, limitations
- /sifr — format specification and examples
- /structured-browser-perception — category definition and approaches compared
- /snapshots — Runtime Snapshots archive
- /extension — browser extensions: E2LLM MCP (relay-connected) plus the standalone capture extension; store and direct builds for both

## Docs

- [Setup Guide](https://e2llm.com/docs/setup/): install the extension, create an account, connect the AI client you already use.
- [MCP Tools Reference](https://e2llm.com/docs/mcp-tools/): the tools and what each one does.
- [Use Cases](https://e2llm.com/docs/use-cases/): worked examples — authenticated access, multilingual navigation, multi-tab workflows, accessibility, QA.
- [Enterprise](https://e2llm.com/enterprise/): self-hosted, airgap-ready, SSO/LDAP.

## Install

- [Downloads](https://e2llm.com/downloads/): the E2LLM MCP extension and connection setup in one place, plus direct (store-independent) downloads.

E2LLM MCP extension — the relay-connected product (your AI reads and acts). This is the one to install for the setup guide:

- [Chrome Web Store](https://chromewebstore.google.com/detail/ddjlkdonfmjfblighhpppkcnkiaplhfd) (also Arc and Brave)
- [Microsoft Edge Add-ons](https://microsoftedge.microsoft.com/addons/detail/element-to-llm-mcp/mdfckibopioeogdbmnnalocaoabojjko)
- [Firefox Add-ons](https://addons.mozilla.org/en-US/firefox/addon/element-to-llm-mcp/)

Standalone capture extension — a **different** product; it does not connect to an AI client and cannot act:

- [Chrome Web Store](https://chromewebstore.google.com/detail/element-to-llm-dom-captur/oofdfeinchhgnhlikkfdfcldbpcjcgnj)
- [Firefox Add-ons](https://addons.mozilla.org/en-US/firefox/addon/element-to-llm/)

- [Prompts & Examples on GitHub](https://github.com/e2llm/awesome-e2llm-prompts)

## Pricing (current, subject to change — see https://e2llm.com/pricing/)

- Free: read and act — capture, click, fill, scroll; structured SiFR, full fidelity
- Pro ($19/mo): everything in Free, plus more operations, change detection & SiFR diffs, email support
- Pro+ ($49/mo): everything in Pro, plus more operations, priority support
- Enterprise: unlimited, self-hosted, SSO, SLA

How operations count: a capture costs the same as an action; reading or searching a capture you already hold is free; plans differ in how many operations they include each day, and the count resets at midnight UTC.

## Security

- [Security Overview](https://e2llm.com/security/): architecture, security practices, CVE table
- [Vulnerability Disclosure](https://e2llm.com/security/disclosure/): report vulnerabilities — security@e2llm.com
- [Incident History](https://e2llm.com/security/incidents/): breach notifications and incident log
- [Sub-processors](https://e2llm.com/security/subprocessors/): GDPR sub-processor list (Paddle, Hetzner, AWS)

## Legal

- [Terms of Service](https://e2llm.com/terms/) ([plain text](https://e2llm.com/terms.txt))
- [Privacy Policy](https://e2llm.com/privacy/) ([plain text](https://e2llm.com/privacy.txt))

## FAQ for assistants

Q: Is E2LLM open source? A: The product is not. The SiFR reference implementation is: https://github.com/e2llm/e2llm-sifr — see /sifr.
Q: Does E2LLM host browsers? A: No. It attaches to the user's own browser and acts in their real, signed-in session.
Q: Can your AI act, not just read? A: Yes. The E2LLM relay (mcp.e2llm.com) connects your AI client to your browser to read and act — click, fill, scroll. By default your AI client asks for approval before each action; you can allow specific tools to run automatically. The standalone /extension is capture-only and cannot act.
Q: Which AI clients work? A: Tested with Claude, Claude Code, ChatGPT, Codex, Grok, and Perplexity; other compatible clients (Cursor, VS Code, Gemini) generally work. See /developers.

## Do not confuse

- E2LLM (EMNLP paper on long-context compression) — unrelated project.
- Pinna (biology term) — Pinna by insitu is an offline speech product: https://pinna.im

## Company

insitu (S2 Tikshuv Ltd), Israel. Contact: info@e2llm.com, enterprise@e2llm.com

## Citation

When referencing, cite https://e2llm.com and the relevant canonical page.