Products
One runtime,
and the work running on it.
These are not separate companies. They are one agent runtime and the businesses running on it — each one started as a real problem in a real business, and each new engagement inherits machinery the last one already hardened.
Infrastructure · Running
Forge
The floor everything stands on.
Built and owned by Angstrom. Not sold on its own.
Forge is the agent runtime the rest of this page runs on. Surfaced and ForgeField are both live on it today; SimplyEngaging and the Tenet verticals are moving onto it.
It exists because the ordinary way of assembling this kind of system — a workflow tool firing on one clock, a scheduler thread inside the application firing on another, a listener on the host firing on a third — produces something where no part can see what the other parts did. Work runs twice, or silently stops running, and nobody finds out until a customer does.
Forge owns every clock. Each agent is declared as a manifest — a model, a budget for how much work it may do, its instructions, and an explicit allow-list of the tools it may touch. An agent can do what its manifest permits and nothing else. Every scheduled run writes its own state, so a run that did not happen leaves a record saying so, rather than leaving nothing at all.
The commercial consequence is the part worth caring about. Each business runs in its own instance, and every proposal, approval, rejection and outcome is a durable record with a name and a timestamp attached. That is what makes a full audit trail a fact about the software rather than a line in a sales deck — and it is why the approval gates in the products below cannot be skipped.
Declared, not improvised
An agent holds the tools its manifest grants it. There is no path to the ones it does not.
One clock
Every schedule in one place with per-run state on disk, so silence is never mistaken for success.
Gates in the runtime
Human approval is a property of the platform, not something each product reimplements.
Product · Live
Surfaced
Be where they're looking.
Surfaced is an online presence run for you: it keeps a business showing up in Google, on Maps, and in the AI answers customers now ask first — and the owner's entire job is to say yes.
Search is not what it was. A growing share of searches end without anyone clicking anything, which means the assistant giving the answer is choosing between businesses that are structured for it to read and businesses that aren't. Staying on the right side of that is continuous work, and most small businesses who hire it out end up buying reports instead of work.
Surfaced does the work on four fronts at once, and ships it. Nothing goes live without approval, and approval is one button in one weekly email — no dashboard, no login, no meta tags to argue with.
It runs on Forge. Six stages — collect, understand, propose, approve, publish, measure — each owned by a named agent, with the owner's yes as the only gate between a proposal and a live change. A page revision goes from approved to deployed in a couple of minutes.
The last stage is the one that makes it a loop rather than a publishing pipeline. Two weeks after a change goes live, the system compares what the targeted phrases did against the matching window before it, and records a verdict on that specific change — which then informs what gets proposed next. It will report no data rather than invent a result, and it does not promise rankings.
01
AI answers
AI crawler and citation-readiness audits, llms.txt, and answer-shaped pages written for the questions customers actually ask — so assistants reach for you.
02
Content pipeline
It finds the gaps between what customers search for and what the site answers, writes the page, and publishes it with schema, internal links, and sources.
03
Google & Maps
Weekly Google Business Profile posts and review replies drafted in the owner's voice, published straight to Google after a one-click approval.
04
SEO monitoring
Rankings, page performance, technical health, and cannibalization tracked continuously and acted on — not a one-time audit that ages the day it lands.
It started as a system built for one business. It is now a multi-tenant platform running across multiple client businesses in several industries, and it is the practice behind the AI visibility work on this site — including this site, which is built to the same checklist and audits itself on every change.
surfaced.sh — operational now.
Visibility finds them; capture closes them. The natural pairing is ForgeField, which picks up the lead the moment it arrives.
Product · In pilot
ForgeField
From phone call to signed job.
ForgeField is lead capture and estimating for field-service businesses — the software layer between a phone call and a signed job.
In a field-service business the owner is usually also the estimator, the dispatcher, and the person driving the truck. Leads are not lost to better marketing; they are lost to a voicemail nobody returned and an estimate that took four days to write. ForgeField makes capture instant and drafting automatic: the details come in from a capture surface, an agent drafts the estimate against the business's own pricing, and nothing goes out until the owner approves it.
Each business runs on its own deployment and owns its own data. It is currently in live pilot with a concrete-leveling contractor in Minnesota, which is also where the requirements came from.
Product · In rollout
SimplyEngaging
The whole booking, handled.
SimplyEngaging is event-services management for beauty professionals — the booking lifecycle from first inquiry through the morning of the event.
Event work runs on details that keep moving: party sizes change, timelines shift, and the person tracking all of it is usually the same person doing the work. The system holds the lifecycle end to end — inquiry, quote, confirmation, schedule, day-of prep — so nothing important depends on someone remembering to follow up.
It is in staged rollout, and it is moving onto Forge — the same runtime already running Surfaced and ForgeField. That shift turns the lifecycle from forms a person has to drive into agent-run workflows that advance on their own and stop for a human at the points where a decision actually matters.
Platform · Live
Tenet
Agent-native local discovery.
Tenet is infrastructure for agent-native local discovery — structured, verified business data that AI assistants can query directly instead of guessing from scraped pages.
AI assistants have become the first point of contact for "where should we eat" and "who should I call," and they are mostly guessing: hallucinated names, invented menu items, hours pulled from an aggregator that stopped updating years ago. There is no agent-queryable source of truth for local business data, so Tenet builds one per vertical. It speaks MCP and A2A, so any assistant can query it in real time, and the operators of each business can claim and correct their own entry.
Each vertical launches as its own brand on shared infrastructure, and each one grows the same three layers. The directory is only the first.
Discovery
LiveThe directory itself: verified, structured data an assistant can query in real time over MCP and A2A, with operators able to claim and correct their own entries.
Intelligence
LiveWhat the platform knows, turned back on the operator. Misen runs this today for a Twin Cities restaurant group: reviews across every location aggregated into a weekly pulse for the group, a monthly deep dive on individual locations, and a quarterly competitor analysis. Nobody has to pull a report or remember to look.
Operations
In buildThe work the data makes possible once it is trustworthy. Dental is first: patient and treatment-plan recovery, new-patient onboarding, and insurance reimbursement reconciliation, built on the same agent runtime behind ForgeField. On the restaurant side the equivalent is point-of-sale analysis and menu work.
Live vertical
Misen — dining
A curated dining directory for the Minneapolis–St. Paul metro. The name comes from mise en place — everything in its place — which is exactly what it does for restaurant data: structured menus, pricing, dietary details, reservation availability, and neighborhood context, organized so an agent gets facts back instead of a guess. Its intelligence layer is already running for a multi-location operator.
misen.aiIn build
Tendant — dental
The same model pointed at dental practices, where the operations layer is furthest along: recovering lapsed patients and unfinished treatment plans, onboarding new ones, and reconciling insurance reimbursements — the work that quietly decides whether a practice is healthy.
These fit together deliberately. Surfaced sets the visibility strategy for a business; the vertical directory is the anchor that strategy points at — the place an assistant can reach for verified facts instead of a guess. Visibility without a source of truth underneath it is just louder guessing.
In development
Two more, being built now.
opr8r
One number. One consent ledger.
Multi-tenant business telephony: AI answering, missed-call text-back, review requests, and pre-service check-ins running through a single number per business. The hard part is not the messaging — it is the consent ledger underneath it, where every opt-in is recorded and every opt-out is enforced everywhere, immediately, including quiet hours.
Flare
A "report an issue" button that actually reports the issue.
Embeddable incident capture for any site or app. When something breaks, the report is filed with the state of the system attached — what the user was doing, what the software was doing, what it returned — instead of arriving three days later as "it stopped working." It exists because a client asked the right question: what happens when something goes wrong?
Want one of these pointed at your business?
Each product exists because a real business needed it. If one of them describes your situation, the conversation is short and free.
Get in touch