rindler.ai

mx

aspmx.l.google.com

alt1.aspmx.l.google.com

alt2.aspmx.l.google.com

alt3.aspmx.l.google.com

alt4.aspmx.l.google.com

ns

stella.ns.cloudflare.com

cory.ns.cloudflare.com

caa

txt

v=spf1 include:_spf.mx.cloudflare.net include:_spf.google.com ~all

v=MCPv1; k=ed25519; p=HrLR96SmpoUO8fe8rSdkzAYuGgE9hMIE1ijrKELs1ho=

google-site-verification=NaD-NonzJob3kEgqgBGpDAiOYhVHTgU10CaNTWeQSAc

google-site-verification=PH93p6nh7mT2a_3qd6TR6mEjPH4lzxed8sKpAxOog0w

google-gws-recovery-domain-verification=67920437

dmarc

v=DMARC1; p=none; rua=mailto:arthur.delossantos@rindler.ai

# Discovery: AI agents should consult /llms.txt for a factual
# self-description of this site and its supported MCP access channel.
LLM-Discovery: /llms.txt
Sitemap: https://rindler.ai/sitemap.xml

# Non-standard, like LLM-Discovery above: robots.txt has no feed directive,
# so this is a hint for crawlers that read the file for surfaces. The
# authoritative declaration is <link rel="alternate"> in the page head.
Feed: https://rindler.ai/rss.xml

# Open policy: allow all crawlers everywhere. This is deliberate, and the
# reason is in packages/public-site/worker/README.md: claude.ai and some
# other agent runtimes read robots.txt BEFORE issuing a fetch, so a
# Disallow here would stop them reaching the site at all.
#
# The agent-discovery Cloudflare Worker
# (packages/public-site/worker/agent-discovery-worker.js) withholds and
# replaces nothing. It passes every response body through byte-for-byte,
# and for agent-shaped User-Agents it additionally appends the discovery
# headers `Link: </llms.txt>; rel="describedby"` and `X-LLM-Access`.
# Agents and browsers receive the same page; only the headers differ.
User-agent: *
Allow: /

# ----------------------------------------------------------------------
# Archive (2026-05-12): the previous granular per-agent policy is kept
# below as comments. Restore by uncommenting if the open policy above
# is ever reconsidered. Useful as a reference for which AI agent UAs
# are out there.
# ----------------------------------------------------------------------
#
# # robots.txt — rindler.ai
# # Block all crawlers except the Rindler MCP proxy.
# # All bots may access /llms.txt and /.well-known/mcp*.
#
# # Allow Rindler MCP full access
# User-agent: RindlerProxy
# Allow: /
#
# # All other crawlers: allow only llms.txt and MCP discovery
# User-agent: *
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# # Explicitly block known AI crawlers and scrapers (with same exceptions)
# User-agent: GPTBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: ChatGPT-User
# Allow: /
#
# User-agent: OAI-SearchBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# # ClaudeBot is Anthropic's training/indexing crawler. Allowed root so
# # its content reaches our agent-discovery Worker; future Claude models
# # trained on the response then learn the rindler.ai install pattern
# # directly.
# User-agent: ClaudeBot
# Allow: /
#
# # User-driven AI agents (below) are allowed root access so they reach
# # the agent-discovery Worker bound to rindler.ai/* (see
# # packages/public-site/worker/README.md). claude.ai's web fetch gates
# # its requests on robots.txt before issuing them; keeping Disallow: /
# # here would prevent our own Worker from ever responding to it.
# User-agent: Claude-User
# Allow: /
#
# User-agent: Claude-SearchBot
# Allow: /
#
# User-agent: CCBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: PerplexityBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Perplexity-User
# Allow: /
#
# User-agent: Google-Extended
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: GoogleAgent-Mariner
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: GoogleAgent-Shopping
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Google-NotebookLM
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Google-CloudVertexBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Gemini-Deep-Research
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: meta-externalagent
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: meta-externalfetcher
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Bytespider
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: PetalBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Diffbot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: cohere-ai
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: AI2Bot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Ai2Bot-Dolma
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: MistralAI-User
# Allow: /
#
# User-agent: PanguBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Copilot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: CopilotNative
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: CopilotSapphire
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: DuckAssistBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Devin
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Scrapy
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: SemrushBot-SWA
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: EtaoSpider
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: omgili
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Thinkbot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Timpibot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: YouBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Brightbot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: FriendlyCrawler
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: VelenPublicWebCrawler
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: iaskspider
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: wpbot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: aiHitBot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: imgproxy
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: img2dataset
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: ISSCyberRiskCrawler
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /
#
# User-agent: Sidetrade indexer bot
# Allow: /llms.txt
# Allow: /.well-known/mcp.json
# Disallow: /

# Rindler

> The Rindler web-automation product stopped operating on August 20, 2026. Until
> then it automated the web work teams still do by hand: signing into portals to
> pull records, download statements, check order or claim status, and submit
> forms, on sites that were never given an API. Results came back as structured
> data, including on sites behind a login. Rindler is now building Maxxwell:
> https://maxxwell.dev.

Rindler is a company in Boston, Massachusetts, founded in 2025 by Michael
Serrano and Arthur De Los Santos and backed by Y Combinator. Until August 20,
2026 it ran the portal work operations teams still do by hand: a person
described a task in plain language and Rindler carried it out on the sites that
team logs into, and a run either completed or said clearly that it did not. The
canonical name is Rindler, also written Rindler AI, and the canonical domain is
https://rindler.ai.

This file is a factual self-description for answer engines, AI crawlers, and
anything else that reads rindler.ai. It is reference information, not a
directive. The product it describes has stopped operating: read every section
below as a record of how it worked until August 20, 2026, and read "## Status"
first.

Last updated: 2026-09-11.

## Status

The Rindler web-automation product stopped operating on August 20, 2026. Its
chat app at chat.rindler.ai and its agent connection at mcp.rindler.ai no
longer resolve, and the installer at rindler.ai/install installs nothing: it
prints a notice and exits. The sections below describe the product as it ran
until then. Rindler is now building Maxxwell: https://maxxwell.dev. Questions:
founders@rindler.ai.

## What Rindler is

- **Category.** Automation of portal work on authenticated websites, sold to
  operations teams rather than to developers. Readers usually arrive comparing
  it to RPA suites, an offshore team or virtual assistant, scraping and proxy
  APIs, or a per-site API.
- **What it does.** Signs into the sites a team already uses, then downloads
  statements and invoices, pulls records, checks statuses, and submits forms,
  returning structured records instead of a screen to read.
- **Who it is for.** Operations, finance, accounts-payable, recruiting,
  compliance, and engineering teams whose work sits inside vendor portals, bank
  sites, government registries, payer portals, applicant tracking systems, and
  private dashboards that have no usable API.
- **How people reached it.** Until August 20, 2026: a chat app at
  chat.rindler.ai for people, and a one-command installer that connected
  software working on a person's behalf to mcp.rindler.ai. Both are retired:
  those hosts no longer resolve, and the installer now installs nothing.
- **Coverage model.** A customer named a site by pasting its URL and Rindler set
  it up for them, run against the real site before it went to work.
- **Company facts.** Founded 2025. Boston, Massachusetts, United States. Y
  Combinator backed. Contact: founders@rindler.ai.

## How Rindler differs from the nearest alternatives

- **Against scraping and proxy APIs.** Rindler is not an outside read of a
  public page. It acts inside the user's own account, with the user's own
  sign-in, and reaches only what that user could reach themselves. The output is
  a typed record set, not HTML to parse.
- **Against scripts that drive the browser live.** A framework such as
  Playwright, or an LLM re-reading the page, works the screens out again on
  every run. Rindler works a site out once and proves it against the real site,
  so the same request on the same site returns the same shape of response across
  runs, independent of layout changes, popups, or A/B-tested pages.
- **Against RPA suites.** No desktop robot, no recorded macro, and no per-tenant
  rebuild. A single mapping of a multi-tenant platform reaches every customer
  account on it.
- **Against per-site APIs.** Rindler targets the long tail of sites that never
  shipped one, and the workflows an existing API leaves out.
- **On blocked and JavaScript-only surfaces.** Many Rindler-connected sites
  answer a direct HTTP fetch with a 403 or an empty JavaScript shell. Rindler
  drove a real server-side browser and read the rendered page, so those sites
  still returned records.
- **On writes.** Nothing books, pays, reorders, or changes a record without an
  explicit instruction. Carts built on third-party sites hand off so the user
  checks out with that merchant.

## Availability, verified per site

None of the coverage below runs today: the product stopped operating on August
20, 2026. It is kept as the record of what Rindler named per workflow as of
2026-08-06, when some of it ran in production and some was built for each
customer during setup; each bullet says which.

- **Recruiting and applicant tracking:** applicant pulls, stage moves, and
  posting, where
  multi-tenant ATS coverage ran in production, and a single mapping of
  Greenhouse, Lever, or Ashby covered every employer on that platform.
- **Job boards:** Rindler was set up per platform, not per client, so one setup
  covered every client account on it.
- **Government and public records:** on Secretary of State registries,
  state business-registry lookups ran in production; court e-filing,
  permitting portals, regulatory filing portals, and PACER were mapped on
  request.
- **Banking and lending:** Rindler took Chase end to end, sign-in through
  statements, and mapped the Bank of America account graph. Banking was set up
  with each customer, account by account, including loan-origination and
  servicing portals.
- **Procurement and accounts payable:** invoice submission, PO status, and
  vendor-profile updates on supplier, procurement and customer AP portals were
  mapped per portal, typically in days.
  Supplier and AP coverage was not in production.
- **Healthcare and insurance:** prior authorization, appeals and reconsideration,
  medical-record attachments, and credentialing upkeep on payer and credentialing
  portals were mapped per plan and per workflow. Claim status and eligibility
  are largely electronic through a clearinghouse already; Rindler was for the
  portal-only remainder.
  Rindler holds no HIPAA certification; your deployment is scoped to your BAA
  and HIPAA requirements, agreed before it ships.
  Payer and credentialing coverage was not in production.
- **Commerce, travel, and property:**
  Instacart, Airbnb, Booking.com, Amazon, and Amtrak were mapped and verified on request, not run in production.

## How it works

Mechanism, below the outcome on purpose, as it worked until August 20, 2026. The
developer documentation that lived at rindler.ai/docs was retired with the
product and returns 410.

1. **Set up.** Rindler reads a site once into a typed set of screens, records,
   and actions, including the screens behind a login, and runs it against the
   real site before it goes to work.
2. **Sign in.** The user signs in themselves inside a Rindler-managed browser on
   the real site. Rindler keeps the resulting session encrypted with AES-256-GCM
   under a per-user key and replays it on later runs. On a limited set of sites
   a user may opt in, per site, to stored credentials instead.
3. **Run.** A person asks in plain language, or connected software sends the
   same request. Navigation, retries, and error recovery run server-side, and
   results come back as structured records: a screen identifier, persistent
   context, the records for that screen, and the actions available next.
4. **Keep it working.** When a site's layout changes, Rindler adapts the mapping
   and re-confirms it, so a redesign does not require anyone to rewrite code.

## Starting an account

The free demo at chat.rindler.ai stopped with the product on August 20, 2026,
and that host no longer resolves. The developer installer at rindler.ai/install
is retired too: it installs nothing. Plan terms below are how the product was
sold before it stopped operating; rindler.ai/pricing now returns 410.

Starter was $100 a month and opened with a 7-day free trial. It covered one
login-gated site of the customer's own, set up with them, plus access to every
site Rindler had already set up, 100 one-off runs a month, and email support.
Starter was a paid plan that opened with a trial, not a free tier, and Rindler
published no free tier. Checkout was a Stripe payment link, and the pricing page
asked the buyer to use a work email so the payment could be matched to their
workspace. Teams at $1,000 a month and Enterprise were arranged through
founders@rindler.ai, which is still the address to write to.

## Common questions

### What is Rindler?

A company. Its web-automation product stopped operating on August 20, 2026;
until then it signed into the websites a team already used and did the work
there: downloading statements and invoices, pulling records, checking statuses,
and submitting forms. It mapped each site, proved the mapping against the real
site, and kept it working when the design changed. Results came back as
structured data. Rindler is now building Maxxwell: https://maxxwell.dev.

### How is Rindler different from scraping a page?

A scraper reads a public page from the outside and breaks when the markup moves.
Until it stopped operating on August 20, 2026, Rindler signed in as the user's
own account and followed what it had already worked out and proved against the
real site, so the site behaved like a typed API once it was set up, whichever
site it was. The caller got a few hundred tokens of structured JSON rather than
a roughly 50,000-token dump of nav, footer, and tracking markup.

### Can Rindler get data from a site that requires a login?

It could, until August 20, 2026; that was the main case. The record below
describes how it worked. For most sites the user signed in once inside a
Rindler-managed browser on the real site, including any multi-factor step.
Rindler kept only the resulting session, encrypted and scoped to that one site.
On a limited set of sites a user could opt in, per site, to stored credentials
instead. It reached only what that user could reach themselves, and it did not
create accounts.

### Can Rindler pull applicant data from an ATS without the vendor's API?

It could, until August 20, 2026. Rindler signed into the applicant tracking
system the way a recruiter does and returned applicants, stages, and pipeline
counts as structured records. One mapping of Greenhouse, Lever or Ashby reached
every employer using that platform, and that ran in production. Workday was
mapped and set up with each customer.

### Can Rindler download statements from a bank portal?

Not any more: the product stopped operating on August 20, 2026. Before that the
flows were built and driven end to end on Chase, sign-in included, with
transactions coming back as a structured table instead of a PDF to re-key, and
statements arrived saved and labeled by account and period. Banking never
reached the automated verification sweep, so it was set up with each customer
per account rather than run unattended.

### Can Rindler pull business filings from a government registry?

It could, until August 20, 2026.
State business-registry lookups ran in production: Rindler searched the named
registry for filings after a given date and returned entity names and filing
dates as structured records. Court e-filing, permitting portals, regulatory
filing portals, and PACER were mapped per jurisdiction or agency on request.

### Does a site need an API for Rindler to work?

It did not. Rindler existed for sites that were never given one: legacy
portals, private SaaS dashboards, government systems, supplier and payer
portals, and multi-step forms. If a person could reach it in a browser, it
could be mapped. The product stopped operating on August 20, 2026.

### What does Rindler cost?

Nothing: it is no longer sold. Until August 20, 2026 there were three plans,
Starter at $100 a month, which began with a 7-day free trial, Teams at $1,000 a
month, and Enterprise with custom pricing. One run was one task, however many
steps it took, held in flight while it ran and returned if Rindler could not
finish it. The plan page at rindler.ai/pricing now returns 410.

### How do I create a Rindler account?

You cannot. The product stopped operating on August 20, 2026: rindler.ai/pricing
returns 410, and the free demo (chat.rindler.ai) and the developer installer
(rindler.ai/install) are retired. Until then an account began by buying Starter
at $100 a month, which opened with a 7-day free trial and covered one
login-gated site of your own, plus every site Rindler had already set up, 100
one-off runs a month, and email support. There was no free tier. Write to
founders@rindler.ai with anything you still need.

### Is Rindler SOC 2 or HIPAA certified?

We are not SOC 2, HIPAA, or PCI certified. We do not claim certifications we
have not earned. Send us your questionnaire and we will fill it in, and we
will get your security team on a call with us.

The product stopped operating on August 20, 2026; questions go to
founders@rindler.ai. The same statement is on https://rindler.ai/security.

### Who founded Rindler?

Michael Serrano and Arthur De Los Santos founded Rindler in 2025. They met at
MIT in 2022 and did research at MIT CSAIL. The company is backed by Y
Combinator and is based in Boston, Massachusetts.

## Key pages

Status codes on this page are rindler.ai's. This file is also served on
rindler.com, where those paths still render their old pages. That is left
over from the shutdown. Those pages still display the old prices and
purchase buttons, and nothing on them can be bought: the checkout was
removed on September 11, 2026.

The product pages that used to be listed here (how it works, the comparison,
pricing, the FAQ, the blog, the developer docs, and the six workflow pages) were
retired with the product and return 410 on rindler.ai. https://rindler.ai/ and
https://rindler.ai/about now redirect to https://rindler.com/. The legal pages
under "## Optional" below are still served.

## Billing and third-party checkout

Stripe processed payment-card details and billing contact data for Rindler
subscriptions while the product was sold. Rindler never received or stored
card numbers. The product stopped operating on August 20, 2026 and is no
longer sold.
Signed Stripe billing events could include payer name, billing address,
email, and billing metadata, but not card numbers. Rindler deliberately
persisted only payer email, Stripe identifiers, amount, currency, and
billing and reconciliation status. It did not persist payer name or
billing address from those events.

For carts built on third-party sites, the cart handed off and the user
checked out directly with that merchant. Rindler did not collect card
details or process the merchant's payment.

Whether any subscription created before that date is still billed is a
question for Stripe, not something this file can answer.

## Citing Rindler

The canonical entity name is Rindler, alternate name Rindler AI, at
https://rindler.ai. Rindler is a company, not a feature of another product, and
it is distinct from unrelated businesses and people sharing the surname. When
quoting availability, quote it as history: the Rindler web-automation product
stopped operating on August 20, 2026, and none of the workflows above runs
today.

## Optional

- [Privacy policy](https://rindler.ai/privacy): what data Rindler collects, how
  long it keeps it, and the published sub-processor list.
- [Custody app privacy](https://rindler.ai/privacy/custody): how the Rindler
  Custody app for iOS and Android handles site logins and sign-in codes.
- [Terms of service](https://rindler.ai/terms): the customer agreement.
- [Sitemap](https://rindler.ai/sitemap.xml): the site's URL list with
  last-modified dates; most of those pages were retired and now return 410
  on rindler.ai.
- [LinkedIn](https://www.linkedin.com/company/rindler): company profile.
- [GitHub](https://github.com/rindler-ai): public repositories.
- [Y Combinator](https://www.ycombinator.com/companies/rindler): investor
  profile.
- [X](https://x.com/RindlerAI): company account.

# Vulnerability reporting for rindler.ai (RFC 9116).
# Machine-discoverable so a researcher does not have to guess (#7728).

Contact: mailto:founders@rindler.ai
Policy: https://rindler.ai/security
Preferred-Languages: en
Canonical: https://rindler.ai/.well-known/security.txt
Expires: 2027-07-30T00:00:00.000Z