We build thesystems thatrun your work.

M.Lighthouse builds automation, AI agents, apps and data products, made to run on their own. One operator, an always-on server, and agents that do the repetitive work every night.

[ Studio ][ Minnesota, since 2020 ]
[ What we build ]

Four kinds of system

Each one is built once and then runs. Each comes with its proof: a live product, a running pipeline, a posted result.

[01] Automation

Automation and AI agents

Scheduled systems that build, check, post and report with no person in the loop. A claim gate reads new copy and bounces anything without a source back to draft. Monitoring sends the result to a phone, so a broken job is a message, not a surprise.

[ Proof ]the nightly loop behind every product further down this page.

A nightly schedule builds and checks a page, a claim gate sends unsourced copy back to draft, the rest is published and reported to a phone.unsourcedNightly scheduleBuild + checkClaim gateBack to draftPublishReport to phone
[ One night's run, schedule to report ]
[02] Apps

Apps, tools and data products

Web apps, internal tools, dashboards, plugins, and data products built from public datasets: government surveys, federal registers, state licence files, court records. Each one is shipped, hosted and kept current by the same nightly jobs.

[ Proof ]fourteen kinds of thing, below.

[ A document tool with paid tiers ]
[ A rate calculator ]
[ A scanner plugin ]
[03] Content

Content and video pipelines

An article becomes a clip, the clip gets posted, and pins and posts are queued and published by jobs. Reels and Shorts are produced from data on a schedule.

[ Proof ]one run of the pipeline: a guide page, the clip cut from it, the pin that went out.

[ 1. Article ]
[ 2. Clip ]
[ 3. Posted ]
[04] Paid media

Paid media, with the reporting automated

Google Ads and Meta, run from a manager account. Every night the results are pulled into one dashboard, so on Monday you open a single page instead of two ad consoles.

[ Proof ]a medical-device client's ad pilot runs on this loop.

Google Ads and Meta results flow into a nightly pull, then one dashboard, then your phone on Monday.Google AdsMetaNightly pullOne dashboardMonday, on your phone
[ The reporting loop, as it runs ]
[ Built and running ]

What we've built

Fourteen kinds of thing, each one live and kept current by scheduled jobs. Drawn here as a kind, not a client.

  • [01]

    Address lookup on a public dataset

    Type an address, get a verdict built from a government survey. Pages regenerate when the source updates.

  • [02]

    Litigation record with alert digests

    Federal court data, rebuilt monthly, with an email digest when a case moves.

  • [03]

    Licensed-provider registry

    A federal register, snapshotted weekly, with a change log.

  • [04]

    State-by-state rules database

    One page per jurisdiction, every claim checked against its source.

  • [05]

    Supply-shortage monitor

    An open federal feed, diffed daily, hundreds of pages kept current.

  • [06]

    PDF tool with paid tiers

    Upload, process, export; free tier, paid plans, no signup to try.

  • [07]

    List and audience audit tool

    Paste two lists, see the difference; a social-follower variant.

  • [08]

    Rate benchmark calculator

    Thousands of data points behind a 3D rate surface and a calculator.

  • [09]

    Practice app with a small store

    A daily practice tool for kids with a paid kit store behind it.

  • [10]

    Repo-to-API packager

    Point it at a repository, get a containerised endpoint.

  • [11]

    Accessibility scanner plugin

    Scans a CMS site, fixes what it can, reports the rest.

  • [12]

    Client ad dashboard

    Ad platforms pulled every night into one report, opened Monday morning.

  • [13]

    Programmatic video channel

    Data in, finished vertical videos out, on a schedule, nobody watching.

  • [14]

    Content-to-social pipeline

    An article becomes a clip and a pin and is posted, by a job.

[ The studio ]

Who

One operator, no account team. M.Lighthouse was started in 2020 in Minnesota after years in healthcare and medical-device marketing. You talk to the person who builds it.

  • 30+Live sites
  • 100+Vercel projects
  • 130+Repositories

The repetitive work is done by agents every night on an always-on server, which is how one person builds and runs this much.

Tools we build with

[ Built with ]
[ Martech ]

[ the stack behind the work; no partnerships implied ]

[ Questions ]

Before you write

Short answers to what people ask first.

  • What does M.Lighthouse build?

    Four kinds of system: automation and AI agents, apps and data products, content and video pipelines, and paid media with the reporting automated. Each one is built once and then runs on a schedule.

  • Do you build websites?

    No. We do not sell website builds or web design. We build apps, agents and data products. Some of them live on the web, but the job is the system behind the page, not the page.

  • What does "runs on its own" mean?

    The system runs on an always-on server on a schedule. It builds, checks, posts and reports with no person in the loop, and when a job breaks you get a message on your phone instead of a surprise.

  • Do you run ads?

    Yes. Google Ads and Meta, run from a manager account. Every night the results are pulled into one dashboard, so on Monday you open a single page instead of two ad consoles.

  • How do we start?

    Write to us. A few lines about the process you repeat, the dataset you have, or the ads account you run is enough. Use the form on this page or email hello@mlighthouse.com, and the person who builds it replies.

[ Contact ]

Tell us what should run on its own

A process you repeat every week, a dataset nobody has turned into a product, an ads account with no reporting. A few lines is enough to start. Or write directly: