Comparison · Claude for Small Business

Claude SMB runs your business.
Setell runs your quote-to-cash.

Generalist agents read the email. A vertical agent already knows how your shop quotes it.

Anthropic shipped 15 agentic workflows and 15 skills covering month-end close, contract review, campaign management, and invoice chasing. That's a real product for the half of small business work that's broadly horizontal. The other half — drafting a quote that sounds like you, remembering Acme always asks net-60, knowing contractors discount differently than print shops — needs a vertical brain. That's Setell. And Setell installs inside Claude.

Join the Setell-MCP waitlist
Claude SMB
Horizontal agent
QuickBooks · PayPal · HubSpotCanva · DocuSignGoogle Workspace · M365
Setell-MCP
Coming summer 2026
Setell
Vertical agent
App · Boxx pipelineGmail · brand-matched draftsMCP · tools, resources, prompts

Honest credit

Anthropic shipped a real product.

Setell is built on Anthropic. We use Claude every day. The May 2026 Claude for Small Business launch is the most credible attempt anyone's made at putting a general-purpose agent in front of the half of the U.S. economy that runs from QuickBooks and email. Five honest credits:

Real workflows, not demos.

Fifteen agentic workflows shipped, not promised. Payroll planning, monthly close, business pulse, campaign management, invoice chasing, margin analysis, month-end prep, tax-season organization, contract review, lead triage, content strategy. These are tasks owners actually do.

Broad, official integrations.

QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, M365 — with partner quotes from leadership at Intuit, HubSpot, and Canva. That's distribution and engineering at a scale no startup matches.

Anthropic-grade engineering.

The safety story, the data-handling defaults (“we don't train on your data by default on Team and Enterprise”), the audit posture — that's all done at a level small SaaS shops will never reach.

The SMB Tour.

Free half-day AI fluency training for 100 SMB leaders per stop. Chicago, Tulsa, Dallas, Hamilton Township, Baton Rouge, Birmingham, Salt Lake City, Baltimore, San Jose, Indianapolis this spring. That's real on-the-ground education for an audience nobody else is teaching.

The horizontal premise is right.

For a lot of routine SMB work — reconciling PayPal against QuickBooks, drafting a campaign, reviewing a contract, prepping the close packet — a horizontal agent with broad integrations is the correct shape.

The structural argument

Where generalist agents structurally can't go for service businesses.

This isn't a roadmap gap. It's a shape gap. Five things a horizontal agent built for 33 million U.S. small businesses cannot ship at platform scale — and the things Setell ships every week because they're the entire product:

01

Per-vertical pricing memory.

Machine shop pricing isn't landscaping pricing isn't HVAC pricing. The same descriptor (“install three”) means three feet, three hours, or three units depending on the vertical. A generalist agent can ask QuickBooks about your cash position. It can't carry the per-shop, per-vertical pricing patterns that make the first quote draft 80% right. Setell does.

02

Per-relationship customer memory.

“Acme always asks net-60.” “Sarah's crew prices T&M, not fixed-bid.” “The Henderson site is always change-order-heavy.” These are per-relationship patterns, not per-customer fields in a CRM. Generalist agents see a fresh conversation every time. Setell remembers — and drafts the next quote with that context already factored in.

03

A revision engine, not a re-draft.

“Change the tile to $3K, drop the demo line, add a $400 dump fee.” That's a structured patch — classified as a direct edit, parameter change, partial change, or contextual change — produces v2 with a full audit trail. A generalist agent would regenerate the quote, losing the line you didn't want changed. The revision engine is a vertical primitive.

04

Brand-matched output that looks like you sent it.

Your logo, your colors, your voice in the email, your PDF template. Customers shouldn't know an agent drafted the quote. A horizontal agent's output is recognizably horizontal-agent shaped. That's a credibility problem in service businesses where the customer is buying you, not your tooling.

05

The Monday-morning brief that sounds like your business.

“Three quotes signed overnight. One stale at Henderson. Two new inbounds drafted — Acme's is net-60 already.” That requires per-vertical, per-relationship, per-customer brain. A generalist agent gives you “cash position, sales trends, pipeline movement” — Claude SMB's actual feature list. Setell gives you the vertical-aware version of that same dashboard.

None of this is a knock on Claude SMB. The horizontal shape is right for horizontal work. Quote-to-cash is vertical work. That's the whole pitch.

Use both

Use Claude SMB for the business.
Use Setell for the deals.

The right SMB stack isn't one agent. It's a horizontal agent for horizontal work and a vertical agent for vertical work — both running on the same Anthropic infrastructure.

Use Claude SMB for
Use Setell for
Month-end close + tax-season prep
Drafting the quote when the email comes in
Contract review (DocuSign)
Revising the quote when the customer pushes back
Campaign management + Canva content
Brand-matched PDFs and email that look like you sent them
Invoice chasing across PayPal + QuickBooks
Two-way QuickBooks sync: customer match, item catalog, payment reconciliation
Margin analysis + business pulse
Per-vertical pricing memory + per-relationship customer memory
Lead triage in HubSpot
Quote-to-cash funnel: sent → viewed → approved → signed → paid
The horizontal half of your business
The vertical half that is your business

You don't pick. You compose.

Feature comparison

Honest about where each one wins.

Most comparison pages claim the new tool wins every row. That's a credibility problem. Below is what each product is actually shaped to do. Three rows go to Claude SMB. Eight rows go to Setell. Three rows are roughly even.

Capability
Claude for Small Business
Setell
Horizontal agentic workflows
close, contracts, campaigns
Yes — 15 shipped
Out of scope
Broad-and-shallow integrations
PayPal, HubSpot, Canva, DocuSign, M365, Workspace
Yes — official partner deals
Some via MCP; not the focus
Anthropic-grade engineering + safety posture
model, infrastructure, audit
Yes — table stakes for them
We build on it
Per-vertical pricing memory
machine shop ≠ landscaping ≠ HVAC
No
Yes — core moat
Per-relationship customer memory
“Acme always asks net-60”
Per-session at best
Yes — persistent, vertical-aware
Structured revision engine
version history + audit trail
Re-drafts the document
Yes — deterministic patches, every version preserved
Brand-matched outbound
your logo, colors, voice, PDF
Generic agent-shaped output
Yes — customers can't tell
Deep two-way QuickBooks sync
customer match, items, invoice, payment recon
Shallow read (cash position, payroll)
Yes — quote-to-cash close-the-loop
Quote-to-cash funnel dashboard
sent → viewed → approved → signed → paid
No
Yes — Pro tier
Configurable autonomy modes
Watch / Trust / Auto per action class
One mode (you-initiated)
Yes — graduates with you
Three peer surfaces
UI, email, MCP — same state
One surface (Cowork chat)
Yes — same state, three doors
In-person SMB education at scale
live tour, 100 leaders per stop
Yes — Claude SMB Tour
No — we're not Anthropic-sized
General-purpose agent for non-quote work
blog post, payroll plan, card recon
Yes — the whole point
No — not the product
Setell-MCP composability
call Setell from your agent stack
Host-side: yes, when shipped
Yes — that's the point

Based on the Anthropic Claude for Small Business announcement (May 2026) and the Setell product as of Q2 2026. We'll update this table when Claude SMB ships vertical-specific skills.

Composition, not competition

Setell installs inside Claude.

The Setell-MCP server is the bet. We're shipping Setell as a primitive that any MCP-aware agent — Claude SMB, Claude Code, Claude.ai, ChatGPT, your bookkeeper's stack — can call. Same vertical brain. Same per-relationship memory. Three doors to the same product:

Door 1

Boxx in the Setell app.

The polished web product. Pipeline, quotes, customers, the morning brief, the funnel dashboard. Operators who want a hand-held experience open this every morning.

Door 2

Setell drafts from Gmail.

Your Gmail is the primary surface. Inbound requests turn into drafts; outbound quotes go through your own Gmail with brand-matched PDFs. No new inbox to check.

Door 3

Setell tools in your agent stack.

Setell-MCP exposes tools, resources, and prompts to any MCP-aware host. Ask Claude SMB to “draft the quote for the Henderson email” — Claude routes the call to Setell, Setell uses your vertical brain to draft it, and the result lands in your Setell app for review. Or your bookkeeper runs the whole quote-to-cash pipeline from Claude Code without ever opening Setell's UI.

Be a primitive Claude calls, not a thing Claude replaces.

Setell strategic posture

Setell-MCP early access opens summer 2026. Join the waitlist →

Who's using both

Three operators. Three stacks. One pattern.

The four ICP segments — app-first, hybrid, agent-first, intermediary-driven — all compose Setell with Claude SMB differently. Three concrete sketches:

Scenario 1

Mike, app-first contractor

(4-person crew, $1.2M/year)
Stack

QuickBooks + Gmail + Setell app + occasional Claude SMB on his laptop.

A typical Tuesday

Mike opens Setell first thing — three quotes drafted overnight from Monday's emails, one stale he's chasing. He reviews and sends two, asks Boxx to add a $400 dump fee on the third. Around 11, he opens Claude SMB to ask “what's my cash position, and which invoices are overdue?” Claude pulls from QuickBooks. He closes Claude and goes back to running his shop.

The split

Setell is his work surface. Claude SMB is the occasional finance assistant. He never feels like he's choosing between them — they own different problems.

Scenario 2

Priya, agent-first machine shop owner

(8-person shop, $3.5M/year)
Stack

Claude Code as her base, Setell-MCP installed, QuickBooks + Gmail wired into both.

A typical Tuesday

Priya never opens the Setell UI. She runs everything from Claude Code. “Draft the quote on the Henderson email — and check our last three jobs with them for pricing reference.” Claude routes the draft call to Setell-MCP. Setell uses her machine-shop pricing memory, applies Henderson's net-60 default, returns a draft. Priya reviews it in the chat, asks for one revision, says “send it.” Claude calls Setell's send-quote tool. Done. Later: “Anything in my QuickBooks I should know about?” Claude SMB handles that.

The split

Setell is the vertical primitive inside her agent stack. Claude SMB is the horizontal primitive inside the same agent stack. She composes both from Claude Code. The Setell UI exists as a backup dashboard.

Scenario 3

Sarah, intermediary bookkeeper

(Acme Bookkeeping, 30 SMB clients)
Stack

Claude Code as her workbench. Setell-MCP installed once. QuickBooks ProAdvisor account. Each client has their own Setell tenant.

A typical Tuesday

Sarah opens Claude Code, switches to her “Henderson Plumbing” project context. “Draft today's three quote replies from Henderson's inbox, using Setell.” Claude routes through Setell-MCP for the Henderson tenant. Drafts appear in Henderson's Setell app for Henderson's owner to approve. Then: “Switch to Acme Welding. Run the same play.” Different tenant, different vertical brain (machine shop pricing patterns), different customer memory. Same Sarah, same Claude Code session. For each client's month-end, Sarah switches to Claude SMB workflows: reconcile QuickBooks, draft the close packet, flag overdue invoices.

The split

Setell is how Sarah's clients close deals. Claude SMB is how Sarah's clients close their books. Sarah configures both once and the agents do the work.

FAQ

Honest answers about the frenemy question.

Are you competing with Anthropic?
No. Setell is built on Anthropic infrastructure and ships Setell-MCP into the Anthropic agent ecosystem. We compete only in the sense that we built a vertical agent — a shape Anthropic's general platform structurally won't take. Different products, same stack.
What happens if Anthropic ships a “Quoting for Contractors” skill?
It's a real risk. Our planning assumption is that vertical skills ship in Q3–Q4 2026. Our mitigation isn't to hope they don't — it's to be already inside the agent ecosystem (Setell-MCP), already deeper than they can be on QuickBooks (two-way sync, item catalog, payment recon), and already carrying per-vertical pricing memory and per-relationship customer memory at a granularity that doesn't generalize to a platform. The vertical brain is the moat. The model is not.
Why can't Claude just do this with the right prompt?
It can do some of it, once. It can't do it 10,000 times with persistent per-shop pricing patterns, per-relationship customer memory, brand-matched outbound, a revision engine that classifies four kinds of patches, and a quote-to-cash funnel dashboard. Those are products, not prompts. Setell's revision engine is pure deterministic code — that pattern doesn't live in a general agent's context window. It lives in a vertical primitive.
Do I need both products?
Most operators benefit from both eventually. The smallest shops (1–3 people) often start with just Setell because quote-to-cash is the bottleneck. Larger shops and intermediary-driven setups pick up both faster — Setell for the deals, Claude SMB for the back office. Neither requires the other.
How does Setell-MCP work?
Setell-MCP is a Model Context Protocol server that exposes Setell's tools (draft a quote, send a quote, revise a quote, create an invoice), resources (your pricing memory, your customer memory, your quote history), and prompts (the vertical brain, packaged) to any MCP-aware host: Claude SMB, Claude Code, Claude.ai, ChatGPT, an intermediary's stack. You install Setell-MCP once, point it at your Setell tenant, and any host can call Setell from inside a chat. It's not shipped yet — early access opens summer 2026. Waitlist below.
What's a vertical agent and why does it matter?
A vertical agent is an agent built for one specific kind of work — in our case, service-business quote-to-cash. It has per-vertical brain (pricing patterns, revision patterns, follow-up cadences specific to machine shops vs landscapers vs HVAC), per-relationship customer memory, and the workflow state machine for that specific cycle. A generalist agent (Claude SMB, ChatGPT) is great at horizontal work — but it composes integrations ad-hoc per session rather than running a real workflow with state. For quote-to-cash, the vertical shape wins because the work is shaped that way.
Is Setell secure enough for my customer data?
Setell is multi-tenant by design, scoped by userId at every API route and database query. We pass through CASA + SOC 2 + ISO 27001 controls. Anthropic processes the AI prompts under their Team/Enterprise terms; we don't train on your data. Per-tenant encryption keys for credentials; cancel-time teardown for Gmail watch lifecycle. Full security details at /security.
What if I'm already on Jobber, Housecall Pro, or QuickBooks?
Setell composes with all three. Keep Jobber/HCP for dispatch, use Setell for quote-to-cash. Setell's QuickBooks integration is deep — customer matching, invoice creation, payment reconciliation, two-way sync — so the loop closes regardless of where your quoting starts.
Comparing to a different tool? Jobber, Housecall Pro, Quotient.

Use Claude SMB.
Use Setell.
Compose them.

Free tier on Setell. No credit card. First AI-drafted quote in five minutes. Setell-MCP early access opens summer 2026.

Join the Setell-MCP waitlist
See all comparisons →