Every capability

The loop, listed. One bill.

The pages with a URL have a URL. This is the list to work from if you are filling in a requirements matrix.

Find

Local businesses, from directories you already trust.

Scout searches Google Places by vertical and metro, with Facebook page signals on what it finds. It never contacts anyone.

Find

106 catalog detectors. A list without a problem is not an opportunity.

Website, SSL, ADA, GBP, hiring, permits, service-gap, AEO. Each detector is a sellable finding.

Find

Bulk scan a list. Deep-diagnose one business.

Radar scores opportunity, models dollars with assumptions, and stamps the fulfillment action that fixes each finding.

Find

Re-scan what you saved. Surface what changed.

Scheduled diffs on the opportunities you kept. Timing is a finding, not a reminder.

Find

Who to hunt, written so a search can answer it.

ICP profiles turn vertical, metro, exclusions, and an offer into a filter Scout can run.

Find

Citation evidence you can attach to a pitch.

The probe engine samples answer engines. Claims carry k/n, model, and stored evidence.

Find

Category × metro rollups, provenance-safe.

Pricing maps, hiring digest, service-gap density, AI share-of-voice. Hidden until preview is on.

Prove

A forensic memo, not an audit checklist.

Headline, modeled dollars, evidence, roadmap, and the DEMO that should already exist.

Prove

Twenty-six branded proof types. Compose once.

Studio, library, schedules, public share tokens. A credibility gate refuses bare dollars.

Prove

The working sample is the subject line.

I noticed this problem, so I already built you a DEMO. Here is the live link.

Prove

Your mark on every report and portal.

Branding profiles with a live cover preview. The marketer owns the artifact.

Win

Signal-named campaigns. Two dollars, never mixed.

DEV is what the SMB recovers per month. FEE is what you charge once. Forecast sums FEE.

Win

Email-first, and the gate cannot be turned off.

Relay writes from the finding. ComplianceGate brands the send. Supervised is the default.

Win

Wedges written from the detector, not a merge field.

No-Website, No-SSL, Permit-Lead, Chair-Hour Loss, Service-Gap. Stop on reply.

Win

Your domain, verified, or nothing leaves.

Your verified sending domain. Unverified is a blocking gate, not a warning.

Win

Replies classified. The sequence stops.

Interested, question, not now, unsubscribe. SLA clocks. Suggested reply is gated.

Win

Two calendars that never share a grid.

Your booking link in outreach. Their customer bookings on the site Forge built. Isolated.

Win

Consent that cannot be pre-checked.

The only path a compliant SMS list can be born. Bundled marketing opt-in is a publish block.

Build

A live preview in the same session as the finding.

Forge opens from the detector’s fulfillment action. DEMO first. DEPLOY is gated.

Build

Inbound only. There is no cold-call path.

A demo number for the pitch. A live line after they pay. Locked on the Autonomy Matrix.

Build

Service × city pages, schema, answer-first copy.

Local SEO and AEO from the same taxonomy that found the gap.

Build

A month on a calendar. Publish is gated.

Muse drafts. You approve. Nothing posts because a model felt confident.

Build

Widgets, quotes, instant responder.

Speed-to-lead you can demo. The same gap Radar scored.

Build

Drafts. Professional review recommended.

ADA, privacy, cookies. We do not pretend this is a law firm.

Build

The collect page has no gating branch.

Echo drafts replies. Public review buttons for every visitor. Private recovery sits alongside.

Run

Ten specialists. Nova routes. Relay is the only agent that sends outreach.

Scout never contacts. Voice and SMS cannot go cold. The roster is the loop, not a chatbot pack.

Run

Type a sentence. It becomes tool calls.

Nova classifies, delegates, and comes back last. Specialists do the domain work.

Run

Supervised by default. Four locks you cannot enable.

Autonomy Matrix per agent, per action. Cold voice, cold SMS, and review gating stay red.

Run

Mission control, playbooks, the approval queue.

The governance plane is live. Overnight dispatch is honest about what has and has not run.

Run

After they pay, the record does not start over.

Health, renewals, upsells from leftover findings, Stripe invoices on your own account, portal.

Run

Plain language. No internal notes. No costs.

Magic-link, one business, results a client can read. Never the marketer’s tooling.

How to read the feature list

Every row below is a URL because a capability without a page is a chip cloud. If you are filling a requirements matrix for an agency evaluation, start here, then open the pages that match the work you already sell.

The bill is one product. There is no public dollar table yet. Feature gates that need a provider key fail closed and say unknown. They do not draw a green check from missing data.

Read down the five motions, not across a vendor bingo card. Find pages hunt and tag. Prove pages memo and brand. Win pages park, write, and book. Build pages DEMO, then wait for DEPLOY. Run pages route, lock, inherit, and show a portal. If a row cannot name the agent that owns it, it is not a feature on this site.

Proposed tier names are Starter, Pro, and Scale. The shape is subscription plus meters. Checkout is off. Intended overage policy is to pause rather than surprise-charge. That is not a live bill. Do not paste a dollar into your matrix and attribute it to us.

What counts as a feature on this site

A feature is a module a marketer can drive: search, diagnose, report, send, build, collect, invoice, portal. A feature is not a synonym for an agent. Nova routes. Specialists do the domain work. Relay is the only agent that sends outreach.

Twenty-eight product pages live under /product. This list is those pages. Blog posts are not features. We have not written the blog yet.

A detector is a feature when it has a registry row, a truth state, and a fulfillment action if it maps to a fix. A vibe is not a feature. A chatbot persona you can rename is not a feature. An overnight playbook that has not run is not a feature you can put in a deck as live.

Public tokens count. /r reports, /b booking, /f forms. Strangers see those without a Clerk seat. Forgetting to mark those routes public once sent every prospect to sign-in. An unauthenticated surface is part of the module, not a footnote.

How search and ICP features start a hunt

Lead sourcing is Scout. Google Places, with Facebook page signals on what it finds, filtered by vertical and metro. The result is a saved business, not a message. Google Places is required per tenant and is the only live discovery source. Each business is saved once per source, keyed to the listing’s own ID.

ICP is the hunt written so a search can answer it. Vertical, metro, exclusions, required signals, and the DEMO the loop is allowed to pre-build. Quality businesses is not a filter. Dentists in Austin with no online booking is a filter. Change the offer, change the ICP. Scout will still search the same directories. LeadScore will rank a different pile.

Signal detectors are the tags that turn a listing into something you can sell. Website, SSL, ADA, GBP, hiring, permits, service-gap, AEO. Cheap tags ride along at search. The rest wait for Radar. Vertical smart-defaults pre-select families that match the category Scout already used.

You can import a list you already paid for. Importing does not make a rented B2B file look like a local finding. If the row has no detectable problem, LeadScore stays honest and the row sits there looking like what it is: a name.

How detectors, Radar, and Watchtower prove a row

Opportunity radar bulk-scans a list or deep-diagnoses one business. Radar scores opportunity, models dollars with assumptions, and stamps the fulfillment action that fixes each finding. Quick, Standard, or Deep are depth knobs, not personality knobs.

Deep diagnostic is the forensic memo. Headline, modeled dollars, evidence, roadmap, and the DEMO that should already exist. This page is the one-business door. Radar is the list door. Both write the same finding objects. Both refuse a bare dollar.

Watchtower re-scans what you saved and surfaces what changed. Scheduled diffs. Timing is a finding, not a reminder. A metro you hunted in January is not the same metro after a storm, a permit, or a rating slide. Waiting on Watchtower is a valid choice. Refreshing the same hunt until the list feels lucky is not.

AI visibility stores k/n citation evidence across answer engines, plus model and window. Market intelligence assembles category and metro rollups with provenance and a coverage count. Snapshots freeze monthly so movement has a before. Intelligence stays hidden until preview is on. A rollup without provenance does not become a pitch sentence.

How reports and branding print proof

Reports are twenty-six branded proof types. Compose once. Studio, library, schedules, public share tokens. A credibility gate refuses bare dollars. A finding with no evidence and no confidence degrades to a could-not-verify note.

Report branding is your mark on the cover, with a live preview. The marketer owns the artifact. Powered by Nerxus is off on reports by default. Sample-first sells the fix. The platform is how you made it. Share analytics stay in your library.

Sample-first is the working sample as the subject line. I noticed this problem, so I already built you a DEMO. Here is the live link. If you cannot attach a live link, wait a day. A pretty preview with no finding is spec work.

Atlas is the agent. The feature is the studio. Do not write Atlas into a requirements cell as if it were a second sender. Atlas can share through the gate. The send rail is still Relay.

How outreach, senders, sequences, and inbox stay gated

Outreach is email-first, and the gate cannot be turned off. Relay writes from the finding. ComplianceGate brands the send. Supervised is the default. Unverified domains are a blocking gate. There is no tenant toggle that sends from a platform domain as if it were yours.

Senders are your domain, verified, or nothing leaves. Your verified sending domain. The browser sees masked status. Platform transactional mail may ride a platform domain when that is configured. Tenant outreach does not.

Sequences are wedges written from the detector, not a merge field. No-Website, No-SSL, Permit-Lead, Chair-Hour Loss, Service-Gap. Stop on reply. Quiet hours and warmup caps still apply. A twelve-step generic drip that ignores the finding is the thing this feature is trying to end.

Inbox classifies interested, question, not now, and unsubscribe. SLA clocks. Suggested reply is gated. The sequence stops. Do not ask support to disable stop-on-reply so you can finish the sequence anyway. That ask is a refuse.

How pipeline, meetings, and forms stay separate from a Forge DEMO

Pipeline is Pulse. Signal-named campaigns. Two dollars, never mixed. DEV is what the SMB recovers per month, modeled. FEE is what you charge once. Forecast sums FEE. If your deck sums DEV and calls it pipeline, you will have a bad Friday.

Meetings are two calendars that never share a grid. Your booking link in outreach. Their customer bookings on the site Forge built. Isolated. You are not injecting a patient, a diner, or a repair job into the shop's book from a sales sequence.

The pipeline stage named Demo is a meeting. DEMO in Forge is a working sample. Keep the words apart in a requirements matrix. Procurement that scores us as 'has demos' without asking which noun you meant will buy the wrong week.

Forms are consent that cannot be pre-checked. The only path a compliant SMS list can be born. Bundled marketing opt-in on a required service line is a publish block. A discovered phone number is not consent. Sent, if you connect a carrier later, is a carrier. It is not a consent ledger.

How websites, SEO, conversion, and compliance ship as DEMO then DEPLOY

Websites are a live preview in the same session as the finding. Forge opens from the detector's fulfillment action. DEMO first. DEPLOY is gated. You can sell a site to a business that already has one if Radar named a gap the current site fails: no booking, no SSL, no service pages, no collect path. You are selling the gap, not a new aesthetic.

Local SEO and AEO are service times city pages, schema, and answer-first copy from the same taxonomy that found the gap. AEO is a page, not a vibe score. The probe stores k/n, model, and evidence. Muse and the SEO tool write from that gap. We do not claim to replace a rank tracker.

Conversion is widgets, quotes, instant responder. Speed-to-lead you can demo. The same gap Radar scored. Content is a month on a calendar. Muse drafts. You approve. Nothing posts because a model felt confident.

Compliance fixers are drafts. ADA, privacy, cookies. Professional review recommended. We do not pretend this is a law firm. A missing privacy policy is a draft with that label. It is not a same-day malpractice product.

How reputation and inbound voice stay locked

Reputation is Echo. The collect page has no gating branch. Public review buttons for every visitor. Private recovery sits alongside. Echo drafts replies. Low-star replies stay held. Yelp and Tripadvisor stay monitor or assisted. Do not tell an owner we posted live if we did not.

Voice is inbound only. A demo number for the pitch. A live line after they pay. Locked on the Autonomy Matrix. There is no cold-call path in the voice tool. If inbound voice is the whole offer you sell, a voice specialist will feel cleaner. Synthflow wins that isolation. Here the line sits inside Find, Prove, Win, Build, Run.

Cold SMS stays Off. Texts need first-party consent born in Forms, timestamped, channel-specific, on a registered A2P 10DLC number. If you type text everyone on the Austin list into the agent screen, Nova refuses.

These locks render disabled with a lock icon everywhere they would otherwise appear, rather than sitting off a list where someone might think they were forgotten. A feature matrix that scores 'SMS yes' because a consented reminder might exist later is a matrix we will fail on purpose.

How clients, portal, and automation sit after they pay

Clients is Steward. After they pay, the record does not start over. Health, renewals, upsells from leftover findings, Stripe invoices on your own account, portal. Health is Healthy, Watch, or At-risk. A client's revenue never appears on a health surface.

Portal is plain language, no internal notes, no costs. Magic-link, one business, results a client can read. Never the marketer's tooling. One shop per token. Isolation is the feature agencies are buying at this surface.

Automation is mission control, playbooks, the approval queue, activity, health monitor, and guardrails. The governance plane is live. Overnight dispatch is honest about what has and has not run. automation_gate() runs at decision time. The kill switch precedes everything. Each dispatch appends an activity row.

Guardrails is Supervised by default and four locks you cannot enable. Autonomy Matrix per agent, per action. Cold voice, cold SMS, and review gating stay red. Email without a verified domain stays blocked. Most restrictive wins among mode, matrix, and playbook.

How the agent screen relates to the feature list

The agent screen is a sentence that becomes tool calls. Nova classifies, delegates, and comes back last. Specialists do the domain work. The roster is ten names: Nova, Scout, Radar, Pulse, Relay, Atlas, Forge, Muse, Echo, Steward. The count is the loop, not a catalog of personalities you can rename.

You cannot invent an eleventh sender. You cannot merge Scout into Relay so a search blasts. Adding a module later adds tools behind an existing name or a new name with a matrix row. It does not add a cold-call persona.

Skills live as SKILL.md plus tools on the specialist. Scripts inside a skill are supposed to be deterministic when the job is scoring, sweeping, or assembling an invoice. Token spend is for prose and for the ambiguous classify fallback. That split is how a week of Scout searches does not become a surprise usage bill you cannot explain.

Agents are not features. Features are the modules the agents drive. If your matrix has a row for 'AI agents' and no row for ComplianceGate, you are scoring the mascot and missing the send.

What a feature does when a provider key is missing

A feature that needs a key fails closed and says unknown. Search cannot run without a per-tenant Google Places key. That key is yours. It lives encrypted in integration secrets. The browser sees a masked status, never the secret.

GBP is OAuth, not a key you paste. If the pill says disconnected, connect it. Live replies still require the credential to be exercised. We will not flip Google on from our side on a shared platform key.

Yelp deepens a slice of cheap detectors and later reputation monitoring. Job boards, permit feeds, and license sources turn on their own detectors. Until they are present, those detectors stay unknown. Unknown is a first-class state. It is better than a green check that came from nowhere.

Designed or connecting includes Stripe on your own account, GBP OAuth, Facebook, Trustpilot, and inbound voice through Vapi. Not connected includes Google and Microsoft free-busy, and SMS reminders. Assisted includes Yelp and Tripadvisor. Unpublished includes platform checkout and plan dollars. Quote those columns. Do not collapse them into 'integrations, yes.'

Who should use the feature list as a requirements matrix

A solo marketer should pick the five column leaders and ignore the rest for a month. Search, radar, outreach, websites, agents. Supervised stays on. One vertical, one metro, one problem you can fulfill by Friday.

A small team should add pipeline, inbox, meetings, and guardrails before they add intelligence. Collision is the size problem. Two humans diagnosing the same row or mailing the same owner is how a small team looks sloppy. Pulse boards and business uuids are the seams. A named human should own Relay, not 'the team.'

An agency filling procurement should add isolation facts: Clerk organization equals tenant, encrypted secrets, portal one business, no white-label sub-account resale in v1. Do not promise a SOC 2 report from this list. Do not promise HIPAA because a medical book asked.

If Clay, Instantly, Lovable, and GoHighLevel already cover your week, this list is asking whether the seam between those logins is the reason a finding dies in a CSV. You can keep those tools. The feature is the uuid that survives the close.

What we will not put on a feature scoreboard

We will not put a green check on cold voice, cold SMS, review gating, or unverified-domain send. We will not put a green check on overnight autonomous dispatch while the runner is the open half. We will not put a green check on white-label sub-account resale in v1.

We will not put a dollar on a plan row. We will not put a lead-per-hour number. We will not put a close rate. We will not put a customer logo. We will not put SOC 2 or HIPAA. We will not put live Yelp reply as if the API were already exercised.

We will not score ourselves as a more mature CRM than GoHighLevel. Steward is the Run motion. It is not their funnel canvas. We will not score ourselves as a broader SEO suite than Semrush. We will not score ourselves as a faster isolated site generator than Lovable.

A scoreboard that only contains the loop is a brochure. A scoreboard that only contains the residuals is a hit piece. Both halves belong on the page you hand procurement. Get started if you want into the next batch. The public site will not invent a seat price so the form looks finished.

How product pages map to Find, Prove, Win, Build, and Run

Find pages are lead sourcing, signals, radar, watchtower, ICP, AI visibility, and intelligence. Prove pages are diagnostic, reports, sample-first, and branding. Win pages are pipeline, outreach, sequences, senders, inbox, meetings, and forms. Build pages are websites, voice, SEO, content, conversion, compliance fixers, and reputation. Run pages are agents, agent screen, guardrails, automation, clients, and portal.

That grouping is how the header is supposed to be read. A chip cloud that lists 'AI, CRM, sites, reviews' hides the seam. The seam is the fulfillment action on the finding and the uuid on the row.

Related links on each product page point at the next beat, not at a blog. The blog set exists as shells. Detectors versus audits, inbound voice, no review gating, DEV versus FEE, modeled dollars, Watchtower timing, service-gap, sell the screenshot, health bands, one free fix. Those essays are not written yet. Do not cite a blog URL as a shipped feature.

If you searched for best local marketing platform features for agencies, the accurate match is a list of modules that share a business row from search through portal, with Relay as the only agent that sends outreach and four locks that do not turn off. Quote the nouns. Do not quote a synonym pack.

How LeadScore and modeled dollars stay on different screens

LeadScore stays on the search screen as a ranking of how sellable the row looks from cheap signals. It is not a prophecy about close rate. It does not print a dollar. A high score means the pile is worth a deeper look. It does not mean the owner will take a meeting.

Modeled dollars show up later, on Radar, labeled modeled, with the assumptions still attached. DEV on Pulse is that same modeled monthly recovery, inherited, never remixed into a second invention. FEE is what you type as the amount you will charge once. Forecast sums FEE. Steward later shows retainers you actually entered.

If two screens ever disagreed about DEV or FEE on that row, the app treats it as a defect. Pulse does not get its own dollar. Inventing a fourth number so a homepage tile looks busy is how marketers stop trusting the roster. A requirements matrix that has one cell for 'AI valuation' will fail this product on purpose.

Health is a different number family. Healthy, Watch, or At-risk. Drivers can include results from deployed assets, payment status, engagement, review velocity, and recency. The band is an internal relationship score. It is never the SMB's revenue. Mixing LeadScore, DEV, FEE, and health into one hero metric is how a demo dies.

How public tokens count as features strangers can open

Public tokens count as features because strangers are not users and still have to see the work. /r is a report. /b is a booking page. /f is a form. Those routes have to be public in the proxy. Forgetting that once turned every share link into a sign-in wall.

A report token is how Atlas leaves the logged-in chrome without making the prospect a waitlist row. Opens stay in your library. We will not mine /r/ traffic into platform outbound. Branding on the cover is yours. Powered by Nerxus stays off by default.

A booking token is their customer book, isolated from your sales calendar. A form token is how consent is born. The checkbox cannot be pre-checked. A bundled marketing opt-in on a required service line is a publish block. Viewers and respondents are not Clerk seats.

Portal magic links are a different public surface after they pay. One business. Plain language. No internal notes. No costs. No sibling shops. If a token ever opened two shops, that is a defect. Isolation is the feature agencies are buying at this surface, not a color on a board.

How a week uses five column leaders without opening every page

A week uses five column leaders by starting at lead sourcing, opening radar on the three you would take, building the DEMO on websites, sending from outreach, and checking the agent roster so you know who ran. That is enough product for month one. Intelligence can wait until preview is on. Muse can wait if you do not sell a content calendar.

Monday you need ICP and Scout. Tuesday you need Radar, diagnostic, and Forge. Wednesday you need senders, sequences, and inbox. Thursday you need meetings if anyone booked. Friday you need the approval queue more than you need market rollups. Skipping Watchtower in week one is allowed. Pretending a stale list is still a finding is not.

You do not need the agent screen to drive those modules by hand. Nova is faster when the sentence is already an ICP. A weak classify asks one clarifying question instead of guessing a send. If you type email these into the screen after a search, Nova routes to Relay and Supervised still waits.

Agencies adding isolation to the week still do not need every Run page on day one. Guardrails and senders come before portal. Portal comes after Closed. Automation's overnight runner stays the open half. Demo the queue. Do not demo a night that has not been spent.

What a feature does when the background queue is down

A feature that depends on a background scan accepts the job and sits when the queue is down. The UI is supposed to tell you that. A spinner that resolves into a guessed list is how marketers learn to distrust the product. Inngest drains the job. No worker means no list.

Scout searches can be accepted and unfinished. Radar bulk scans can be accepted and unfinished. Watchtower diffs can be accepted and unfinished. The marketing site will not pretend a background scan is instant. It is a job. Jobs need a worker.

Sends still wait on ComplianceGate even when the queue is healthy. A queued draft is not a send. Activity should name the specialist and the gate decision. An orchestrator that hides the worker is how you lose the audit. Nova comes back last with what ran, what is waiting, and what the gate refused.

If you are filling a matrix that requires 'real-time enrichment,' you will score us down, and you should. Unknown and sitting are first-class states. They are better than a green check that came from nowhere. Settings, Connections is where missing keys go away. The worker is where sitting jobs go away. A requirements cell that cannot accept sitting will also reject Watchtower diffs and Radar bulk scans, because those are jobs too.

Questions

Is every detector a feature I can sell on day one?

No. Scout rides cheap tags at search. Radar runs deeper families. Adapters without a key stay unknown. Sell the finding you can fulfill this week. A missing privacy policy is a labeled draft, not a same-day malpractice offer.

Does a feature page mean the module is priced?

No. There is no public dollar table. Subscription plus meters is the shape. Checkout is off. A page is a URL for a module, not a SKU.

Can I turn on cold SMS or review gating from a feature flag?

No. Those rows are locked on the Autonomy Matrix. There is no hidden tenant toggle. Forms consent is the only SMS on-ramp. Public review buttons stay public.

Are agents the same as features?

No. A feature is a module you drive. An agent owns a slice of the loop. Nova routes. Relay is the only agent that sends outreach. Scoring 'AI agents' without scoring ComplianceGate misses the send.

Why do some integrations say unknown?

Unknown means the key or fetch is missing. It is a first-class state. Google Places is required for search. Other sources stay optional. A green check with no artifact is a defect.

Is the overnight playbook runner a feature I can demo as live?

Governance is live. The overnight runner is the open half. Demo the approval queue and the locks. Do not demo a night that has not been spent.

Find the leak. Ship the work. Keep the account.

Tell us your vertical, your metro, and what you can fulfill. Supervised is the default. Plans start at $99/mo.

Get started See pricing

14-day trial. Card required, converts on day 15. Full table on /pricing.