Tool · Accounting firm ops · TaxDome · Oct 2026

TaxDome AI review (Oct 2026): AI document checklists from prior-year returns, auto-tagging, Juno return prep, the Atlas agent beta, and MCP connections for tax firms

TaxDome is an all-in-one practice-management system with a client portal, used by tax, bookkeeping, and accounting firms for organizers, document collection, e-signatures, billing, and workflow. Its AI today sorts and renames client uploads and builds document checklists from last year's return; Juno turns those documents into a draft return; and Atlas (an AI operations agent) plus MCP and API connections are in beta. This page is about using them while the preparer of record still owns every checklist, field, and message that reaches a client. Each section has a Try this.

Pair with Karbon AI (collaborative email and workflow) · Also new: Karbon AI · Blue J · Oct 2026 · taxdome.com

What this TaxDome page covers

This page is for a firm owner, tax manager, or firm administrator at a tax or bookkeeping practice that runs on TaxDome or is evaluating it. It is not tax advice, and nothing here replaces a CPA or EA reviewing the return.

TaxDome's AI page (read 21 Oct 2026) says document tagging and renaming are available today and that “this fall” TaxDome AI expands into Atlas, open MCP connections for Claude and ChatGPT, and a public API. The Summer Update recap said those three would be in beta over the summer and “available to all firms at the end of Q3 2026”; as of 21 Oct the AI page still says “coming soon” and points to a waitlist, and the MCP help article is labeled Beta. Treat Atlas, MCP, and the API as beta until your account shows them.

Ground rules: the preparer of record checks every AI checklist before a client sees it; a tax manager spot-checks AI tags before staff rely on them; nothing an agent drafts reaches a client until a named person has read it. TaxDome says AI access is opt-in, scoped to each user's permissions, and that firm data is not used to train public or third-party models.

Simple TaxDome AI brand mark diagram
A simple mark we drew for this page. Where we couldn’t get a clean product screenshot, we used simple diagrams of the workflow instead.

Major features (with how someone uses them)

Examples at a glance: each feature has a concrete example from a real job plus a short Try this you can run this week.

1. AI document checklists from prior-year returns (primary)

What it is: TaxDome's checklist assistant builds a document checklist from a client's prior-year return, and as the client uploads, documents are matched to checklist items so you can see progress (Summer Update recap). TaxDome says more than 280,000 checklists were completed during the beta season, and that rollover, several checklists per client, and bulk generation are coming next season.

Last year's return knows last year. A closed brokerage account, a new rental, or a new K-1 changes the list.

Concrete example (preparer of record at a six-person tax practice) The preparer generates checklists for ten returning clients. Seven are right. Two ask for a 1099-INT from an account the clients closed last year, and one misses a new K-1 the client mentioned in an email. She edits all three before sending and adds a line to the intake email asking what changed this year.
Generate, compare with last year, edit, then send
Simple workflow diagram: generate, compare with last year, edit, then send.
Try this

Primary workflow:

Generate AI document checklists from prior-year returns for ten returning clients. Have the preparer of record compare each checklist with last year's return before it goes to the client.

2. Auto-tagging and renaming of client uploads

What it is: With TaxDome AI turned on in Firm settings, client documents are automatically tagged (for example “W-2”), categorized, and renamed, so a file called 4353458223.pdf becomes something staff can find (TaxDome AI). The plan list says TaxDome AI delivers “organized workpapers automatically” on every plan (pricing plans).

A wrong tag hides a document. A 1099-K tagged as a 1099-MISC can sit unseen until review.

Concrete example (tax manager at an eight-person CPA firm) After a week of auto-tagging, the tax manager opens fifty tagged uploads. Forty-seven are right; two brokerage statements are tagged as bank statements and one photo of a W-2 is tagged “Other.” She asks staff to sort by “Other” every morning and keeps a short list of tags to double-check.
Tag, spot-check fifty, fix, then rely
Simple workflow diagram: tag, spot-check fifty, fix, then rely.
Try this

Spot-check first:

Turn on AI tagging for one week of uploads. Have the tax manager spot-check fifty tagged files against the documents before staff rely on the tags.

3. Juno return prep from TaxDome documents

What it is: Juno is a separate tax-prep company TaxDome integrates with: you send client documents from TaxDome, Juno extracts the data and drafts the return from current documents and prior-year data, and your team reviews before export to ProConnect, Drake, Lacerte, UltraTax, or CCH Axcess (ProSeries and ProSystem fx listed as coming soon) (TaxDome help). TaxDome and Juno cite “99% extraction accuracy” — their claim. Juno is billed by Juno per return, with a seven-day free trial of up to five returns.

Extraction errors are quiet: a transposed box number or a missed state line flows straight into the draft.

Concrete example (preparer at a four-person EA practice) The preparer sends five closed returns' documents to Juno and compares every field with the filed returns. Juno matches almost everything; it reads one corrected 1099 as the original and misses a second state's withholding on a W-2. She logs both as “always check” items for the season.
Send, compare every field, log misses, then use
Simple workflow diagram: send, compare every field, log misses, then use.
Try this

Compare before season:

Send five closed returns' documents to Juno. Have the preparer of record compare every extracted field with the filed return before using it in season.

4. Atlas AI operations agent (beta)

What it is: Atlas, per TaxDome's AI page, takes a request such as “draft a payment reminder for every overdue client,” plans the steps, pulls from practice data, and shows a log of every action before anything goes out. It ships with recommended agents (pipeline digests, new client setup) and can turn a plain-English description into a saved agent that runs on demand, on a schedule, or on an event, and “waits for your approval before anything sends.” As of 21 Oct 2026 the page still marks Atlas “coming soon” with a beta waitlist.

A reminder sent to the wrong client, or one who paid by check yesterday, costs goodwill. The action log is where you catch it.

Concrete example (firm owner at a five-person bookkeeping and tax firm) In the beta, the owner asks Atlas for payment reminders to every overdue client. It drafts nineteen. Two clients paid by check that was recorded as an offline payment after the invoice list was pulled. She removes both, sends seventeen, and schedules the agent only after a second week with no mistakes.
Request, read the log, remove errors, then send
Simple workflow diagram: request, read the log, remove errors, then send.
Try this

Read the action log:

If your firm has Atlas, ask it to draft payment reminders for every overdue client. Have the firm owner read each reminder and the action log before anything is sent.

5. MCP connections and public API (beta)

What it is: TaxDome's API & MCP help article (Beta) explains that both are off until a firm owner or admin turns them on, each connection acts as the user who made it, and each has its own access level: Read only, Manage, Full access, or a custom mix. Your assistant reads without asking but asks before anything that deletes data or reaches a client. The article warns there is no test mode: emails, texts, invoices, and engagement letters triggered through MCP or the API really send, so test on a test client first.

“Move this job to the next stage” can fire that stage's automations, including a real invoice to a real client.

Concrete example (firm administrator at a 15-person accounting firm) The administrator connects one assistant with read-only access and asks for overdue jobs by stage; the answers match the pipeline. Before granting Manage access she creates a test client with her own email, moves a test job, and sees an engagement-letter email arrive — exactly what would have reached a client. She documents which stages send email before widening access.
Read-only first, test client, then widen
Simple workflow diagram: read-only first, test client, then widen.
Try this

Test client first:

Connect one AI assistant with read-only access. Have the firm administrator test it on a test client and confirm nothing reached a real client before widening access.

6. AI-driven analytics, Work Forecaster, and Tax Return Delivery

What it is: Pro and Business include AI-driven analytics (plan list). The Summer Update recap (30 Jun 2026) describes Work Forecaster, which flags jobs trending toward a missed deadline, as “coming to beta soon” (the pricing page still says coming soon), and Tax Return Delivery, which presents a finished return with a video walkthrough, highlighted sections, e-signature, and payment in one flow. TaxDome says firms delivered over 120,000 returns this way last season with a 90%+ first-time completion rate, and that AI help placing signature fields and filling payment details is coming.

A forecast of “at risk” is a guess from pipeline data. Jobs parked in the wrong stage make the forecast wrong.

Concrete example (tax manager at a 12-person tax and bookkeeping firm) In March the tax manager asks the analytics which extension-season jobs look at risk. It lists fourteen. Ten are real; four are sitting in a stage nobody uses any more. She cleans up the pipeline stages and reassigns two preparers based on the corrected list.
Illustrative at-risk job list beside what the tax manager checks
Our sketch of an at-risk job check. Not a TaxDome screenshot — the layout is ours.
Try this

Check stages first:

Ask the AI analytics which jobs are at risk this week. Have the tax manager check each against the pipeline before reassigning any work.

Pricing context (October 2026)

TaxDome prints per-seat prices billed upfront (pricing page; help center, checked 21 Oct 2026). Per seat per year on a 1-, 2-, or 3-year term: Essentials $800 / $750 / $700 (one user only), Pro $1,000 / $950 / $900, Business $1,200 / $1,150 / $1,100. A three-year Pro seat at $900 a year is $2,700 upfront for the term.

Temporary seats: Pro monthly seats at $100 per month; Business seasonal seats at $500 per four-month term. Firms with 25+ seats are quoted. A white-labeled client app is a paid add-on with no printed price. AI-driven analytics are listed on Pro and Business; TaxDome has not announced a separate price for Atlas or MCP.

Juno is billed separately by Juno: one flat price per return that falls as volume rises, set on a slider at juno.tax/pricing (no single figure printed on the page), with a seven-day free trial of up to five returns. TaxDome's own pricing page also shows a cost-comparison calculator with other vendors' prices; those are TaxDome's figures, not ours.

Uncertainty: We did not test TaxDome or Juno. Features, dates, and prices come from TaxDome's pricing page, help center, AI page, and blog (read on 21 Oct 2026 through a reader because taxdome.com blocks plain scripted requests) and from juno.tax. Atlas, MCP, and the API are beta and may not be in your account yet; Work Forecaster is coming to beta. Accuracy, completion, and checklist figures are TaxDome's or Juno's claims.

When it’s useful (honest take)

Useful when your firm lives in tax season, clients upload through a portal, intake and document chasing eat staff time, and a named preparer of record will check every AI checklist and extracted field.

Less useful when your firm mainly runs collaborative email and recurring bookkeeping work (Karbon AI), you need cited answers to tax questions (Blue J), or you are managing your own company's books (QuickBooks AI).

Paying for TaxDome AI and a few other AI tools? The $9 AI toolkit spreadsheet lists every AI subscription you have, what each one costs, and how fast it pays for itself, so you can see which ones are worth the money.

Want someone to do it for you? AI toolkit audit — $19

Related options (educational)

Where accounting firm AI jobs can live (Oct 2026)
NeedBetter homeWhy
Client portal, AI checklists, auto-tagging, tax workflowTaxDome AI (this page)$800–$1,200 per seat/year (1-year)
Collaborative email, workflow, close checksKarbon AI$59 / $89 per user/month annual
Cited tax research and memosBlue J$1,498/year Individual
A small business's own booksQuickBooks AIDifferent job
Sanitized client-email drafts in general chatClaude for financeNo client data

FAQ

What AI does TaxDome have in October 2026?

Auto-tagging and renaming of client uploads and AI checklists built from prior-year returns are available now. Atlas, MCP connections, and the public API are in beta, and TaxDome's AI page still lists them as coming this fall.

How much does TaxDome cost in October 2026?

Per seat per year, billed upfront: Essentials $800, Pro $1,000, Business $1,200 on a one-year term, with lower annual rates on two- and three-year terms. Pro monthly seats are $100 a month; Business seasonal seats are $500 for four months.

Is Juno part of TaxDome?

No. Juno is a separate tax-prep company that integrates with TaxDome and bills per return. TaxDome sends client documents to Juno, which drafts the return for your team to review before export to your tax software.

Can an AI assistant change data in TaxDome?

Through MCP or the API it can, within the access level you grant. Both are off until an owner or admin turns them on, and TaxDome warns there is no test mode, so test on a test client first.

Who should check TaxDome AI output?

The preparer of record for every checklist and Juno field; the tax manager for auto-tags and at-risk lists; the firm owner or a named manager for every Atlas action that reaches a client.

Optional next step (if seats overlap)

If you pay for TaxDome seats, a per-return prep tool, and a general chat seat that all touch the same client files, that overlap is worth mapping. Map the overlap with the $19 AI toolkit audit or the $9 AI toolkit spreadsheet. Email Audit $19 or Toolkit spreadsheet $9.

Checkout is live via Stripe. After you pay, we email delivery (spreadsheet) or intake next steps (audit).

Related: Karbon AI · Blue J · QuickBooks AI (client books) · Xero AI · FloQast (in-house close) · Claude for finance · ChatGPT for finance · Microsoft 365 Copilot · Claude · Three checks before another AI seat · ops hub

Affiliate disclosure: Some outbound product links may be affiliate links. We may earn a commission if you subscribe. Prices and features change — always verify on the vendor site. Full policy: disclosure.

Next steps