Lentago Labs

The pledge — We will never host your systems for you. You'll own every piece, we'll show your people how to run it, and firing us is a runbook.

_{Not a slogan — the standing delivery model, written down: ADR-0007 · client-owned delivery, no multi-tenant SaaS.}

Heard the enterprise words but never seen them done small? The asclepias glossary translates CAB, CMDB, PIR and the rest into what we actually do here — and what you can do too.

_{Four things we deliver into estates you own. Each one is already running in the open, and each links to a live receipt: a public repo you can read, fork, and run today.}

_{Not a kit? Audits, migrations, incident response, pipeline hardening — sized to your constraints, delivered into your estate under the same pledge. Free for nonprofits and volunteer-run orgs; everyone else, ask. Get in touch.}

Everything is code. Every change is a pull request — a proposed change that is checked before it lands. Merges apply automatically. A self-hosted fleet of AI coding agents does directed work. Humans own the policy and direct every change. Every directed pull request gets an AI review. Opted-in scheduled routines merge on their deterministic gate alone. That is the whole operating model: nothing here changes except through a pull request, and the merged PR is the change record. The merge promise spells out who approves what.

_{This is not a sandbox of toy YAML. Merge a PR in one of these repos and a live surface moves:}

_{Emphatically free-tier, wherever possible. When a service offers a free tier, that's the one we run — caps and retention windows are treated as real operating constraints to be managed, not something to buy past.}

_{Cloud & containers}

_{Bare metal & virtualization}

_{Infrastructure as code}

_{CI/CD & supply chain}

_{Observability & on-call}

_{Six promises every shop needs, plus three for when you need them. Each part below sits under one promise, and its codename is the part's name. Named systems, honest parts. Each one splits a core that doesn't care where it runs from the pieces specific to us, so the parts specific to us swap out for yours. The current build is always the first client — a working reference, not a finished product. The codenames are New England native plants.}

Your site and giving page — a campaign page you own outright, with a donate button that goes straight to your payment processor.
monarda	Campaign page — a site template, a one-page intake, and a timed dry-run, deploying into the client's own GitHub Pages or AWS. Yours, not ours.
Your records and your proof — minutes, policies, and filings in one vault, the posting rules beside them, and a public board that shows what went up on time. When the board, the auditor, the funder, or the insurer asks, you send a link.
uvularia	Records and deadlines — your public records as plain files with the posting rules next to them; merge and the records publish, a public "Is it posted?" board updates, and a receipt is stamped. Live demo board for a fictional land trust, published by the vault itself; a branded site is an optional second rung.
Your machines and logins — every laptop set up the same way, every login with a named owner, and leaving is a pull request, a proposed change someone else reviews before it lands.
kalmia	Machine setup — turns a fresh Linux install into a fully set-up work machine with one command, and owns every virtual machine and container on our own hardware. Running it again is always safe.
Your alarms and status — know when something breaks before the executive director does, and a file that says who gets the call.
drosera	Alerts and status — what your systems are doing right now, on dashboards anyone can read. One small collector per machine; every dashboard saved as code. If it isn't in the repo, it doesn't exist.
betula	Capture and keep — a complete, searchable record of what happens on a network, kept longer than the vendor keeps it. Our firewall's logs go to a free tier, searchable at $0 a month.
Your help desk and runbooks — requests land in a list, answers come from runbooks you wrote once, and the front desk answers at 9 pm.
mitchella	Front desk — a chat assistant that checks live state before answering from the docs, and drafts a ticket for a human when it can't.
Your data and your exit — every export you'd need to leave a vendor, on a schedule, into a repo you own, with a restore you've drilled. Today, betula (above) keeps the logs longer than the vendor does, and our own lock-in ledger shows what leaving each of our vendors would cost; the scheduled exports and the restore drill are on the roadmap (#123).
When you need them: your building — sensors, and a lobby screen that says whether you're open.
epigaea	Building sensors (temperatures, doors, power) — a production Home Assistant, the open-source smart-home hub, kept entirely as plain-text config (declarative YAML): every change is checked before it lands and rolls back if it fails. The base for a prospective kit for buildings and cold storage (refrigeration that must stay cold) (#129).
brasenia	Lobby screen — one always-on screen whose content is chosen by a rubric, not a remote: panes surface when something needs eyes and sink when it resolves. Roku and Chromecast today.
When you need them: cloud you own — a real AWS environment in your account, as code.
solidago	Cloud you own — our AWS setup, written entirely as code: network, servers, database, web firewall, encryption keys. Serves lentago.dev and three more live sites.
osmunda	Cloud you own — a standing cluster on our own hardware, plus a cloud cluster that exists only for the hours a job needs it. The cluster pulls each approved change itself; nothing pushes to it.
When you need them: AI with receipts — agents that open pull requests and never merge.
claytonia	AI with receipts — a small pool that works unattended on our own hardware. Drop a job, a worker does it on a fresh copy of the code and proposes the change for review. It can't approve its own work; a person always decides.

_{Under all of it, the guide: asclepias (vol. 1, how it all works, with labs) and lupinus (vol. 2, pick a part and run it from an ops vault you own). The pledge above covers every part.}

Three doors, depending on what you came for.

You run a nonprofit's tech and want something you can use today

1. The picker — start from what you need, not from what we built. Every row says what it costs per month and how ready it is.
2. Your first kit — a campaign site with an optional donate button, live in about an hour, deployed into your GitHub account, for free.
3. The pledge (above) — you own every piece, we show your people how, firing us is a runbook.
4. Stuck? chris@lentago.dev. No invoice. You'll hear back inside a day on weekdays.

You want to see it working before you trust it

1. Pick a product repo above and read its 🛠️ Make a change yourself section. Every vector links to a real merged PR.
2. Run a lab against our estate — they start with a first pull request and ladder up to breaking something on purpose (that last one is scheduled with us, not self-serve). A free GitHub account is all you need.
3. Mention @claude on any issue or PR in a product repo and watch the agent fleet respond.

You want to kick the tires on us

1. solidago — the reference AWS platform, 100% Terraform.
2. claytonia — the self-hosted agent fleet that does the directed work, and never merges.
3. Incident register — our post-mortems, published verbatim, including the embarrassing ones.
4. Lock-in ledger — every vendor we depend on, scored on how hard it would be to leave. The receipt behind "firing us is a runbook."

_{Regenerated weekly from the repos themselves — we practice what we publish.}

• Fleet report — open issues by repo, a 30-day activity snapshot, and a code census that counts the CLAUDE.md-family instruction files as natural-language code.
• Epics report — what the open issues add up to: each tracking issue with its sub-issue progress, the month's merged PRs attributed to it, and the work no epic describes.
• Language census — the canonical all-languages breakdown.
• Incident register — post-mortems from running our own estate, with what broke, what did not, and the governance lessons.
• Lock-in ledger — our own vendor dependencies, each scored on export fidelity, format openness, custody, and a documented exit. The receipt behind "firing us is a runbook."