Up to 40x more usage than Claude and ChatGPT
EBITDAI natively runs on DeepSeek, Kimi, Gemini and Meta Muse, allowing best-in-class performance at much higher usage limits than Claude and OpenAI. For the same price as their lowest tier subscriptions, EBITDAI delivers enough usage to handle most workloads.
From $5/month · No 5-hour windows, no weekly lockouts · Excel on Windows, Mac, and the web
45 seconds: what EBITDAI does, start to finish.
One agent in the Excel sidebar. It works on the workbook you have open, cell by cell.
Ask in plain English. The agent reads your sheets, writes formulas, formats, and builds charts and pivot tables.
Read moreThree-statement models, debt schedules, working capital and SaaS ARR, built to a written modeling standard.
Read moreLBO and DCF work: debt tranches, returns and projections built from a portfolio company's books.
Read moreRolling models for several clients, with QuickBooks, Campfire and Stripe data pulled into the workbook.
Read moreEBITDAI is bringing the same AI models and workflows to Google Sheets as a Google Workspace add-on: the identical agent, running in a sidebar beside your spreadsheet. It reads and writes the sheet you have open, builds and edits models the same way it does in Excel, and signs in with the same account.
The Google Sheets add-on is built and working end to end. It is not yet published in the Google Workspace Marketplace; until then, EBITDAI runs in Excel on Windows, Mac, and the web.
EBITDAI runs on DeepSeek, Kimi, Gemini and Meta Muse, whose tokens cost roughly 4-75x less than Claude API rates. This translates to a generous monthly allowance included in each subscription, as well as the ability to pay as you go without breaking the bank. Prefer flat-rate all-day modeling? EBITDAI is also the only add-in that runs on a Kimi Code subscription. Combined with a Pro subscription, this produces the single best value of any Excel agent. The bars below show what each plan's included allowance buys in complete model builds on the model each bar names: the flat-rate Kimi Code bar on top, then Meta Muse Spark 1.3, the cheapest model in the plans, or DeepSeek V4.1 Flash. See the math: DeepSeek in Excel · Kimi in Excel.
Included usage by subscription: complete model builds per month
The top bar is the flat-rate path: Pro's $15/month plus a $19/month Kimi Code subscription, $34 all-in, buys up to 4,400 builds on Kimi k3. Under it are the two Pro bars, the same monthly allowance, 7.5x more usage than lite, priced two ways: on Meta Muse Spark 1.3, the cheapest model in the plans, it buys 1,354 builds, and on DeepSeek V4.1 Flash it buys 663. Lite's allowance, 100x more usage than free, buys 180 on Meta Muse Spark 1.3, the only model Lite runs, and Claude Pro is shown at its best case of 85. Those are the chart's own numbers, and they are no longer turned into a multiple against Claude or ChatGPT: the old per-dollar multiples were measured against a monthly allowance this pricing retired, so they went with it. Every EBITDAI plan is one simple monthly pool with no 5-hour windows or weekly lockouts, so a single deadline sprint can use as much of it as you like, and on Pro your own DeepSeek key builds about 880 complete models for $10, about a cent each, working the minute you create the key.
Built from our published benchmark: one complete model build costs $0.0113 at published DeepSeek V4.1 Flash off-peak prices and $0.0055 on Meta Muse Spark 1.3's Contributor rate card, both priced 2026-09-10; Claude Pro and ChatGPT Plus figures are third-party estimates. Every EBITDAI bar is a plan's monthly allowance divided by that per-build cost, at the allowances the plans publish today: Pro at 7.5x more usage than lite, and Lite at 100x more usage than free. The 4,400 bar is the Kimi Code path, where EBITDAI's $15/month carries the plan and you supply a $19/month Kimi Code account on top of it, $34 all-in; on the Free, Lite and Pro plans by themselves there is no second AI account or bill at all. The Kimi Code figure is a third-party estimate, on the benchmark's one-window-per-working-day math, since Kimi publishes no exact quota.
No opaque quotas, no rationing math, no surprise walls. You always know where you stand: your remaining usage shows right in the pane.
Your usage is one monthly pool. Work in bursts, model all night before a deadline, spend it however you like. Nothing cuts you off mid-task and tells you to come back later.
Pro's $15/month allowance builds 1,354 complete models a month on Meta Muse Spark 1.3, the cheapest model in the plans, and 663 on DeepSeek V4.1 Flash, and Lite's allowance, 100x more usage than free, builds 180 on Meta Muse Spark 1.3, the only model Lite runs. For scale, that is against about 40 for a $20 ChatGPT Plus plan and 85 in Claude Pro's best case, dedicated to Excel and not shared with chat and coding. It refreshes every month, automatically.
Buy extra usage on the spot ($5, $10, or $25, right in the pane) at a far lower price than Claude's extra usage. It never expires, and on Claude the alternative is a $100/month plan.
All five tools work agentically in a spreadsheet. The differences are what a token costs, how hard the meter caps you, and the modeling layer on top.
| EBITDAI | Claude for Excel | ChatGPT for Excel | Shortcut AI | Copilot in Excel | |
|---|---|---|---|---|---|
| Price | $5 to $15/mo | $20 to $200/mo | $20 to $200/mo | $125/mo billed monthly | $30/user/mo |
| Approximate cost per million tokens (in/out) | Meta Muse (Contributor tier): $0.10 / $0.20. Meta Muse (standard tier): $1.25 / $4.25. DeepSeek: $0.22 / $0.66. Kimi k3: $3 / $15. Gemini 3.8 Flash: $0.75 / $3.75. Pro + Kimi Code subscription: ~$1.13 | $20 Pro plan: ~$9 / $15 | GPT-5.6 API: ~$3 / $15 | Not disclosed | ~$4 / $20† |
| Extra usage cost per million tokens (in/out) | Meta Muse (Contributor tier): $0.10 / $0.20. Meta Muse (standard tier): $1.25 / $4.25. DeepSeek: $0.60 / $1.80. Kimi k3: $6 / $30. Gemini 3.8 Flash: $0.75 / $3.75 | $10 / $50 | ~$3 / $15 | $0.10 / credit | No self-serve top-up |
| Usage limits | Simple monthly quota | Monthly, weekly, 5-hour, and per-model limits, shared with all Claude use | Weekly caps on Thinking models; separate timers for files, images, and agent use | 1,000 credits/mo at 2 to 15 per message | Capacity- and credit-based, not published; users report daily limits |
| Modeling benchmark score | Kimi k3: 100% (30/30 tasks). Gemini 3.8 Flash: 93% (28/30). DeepSeek: 87% (26/30). Meta Muse Spark 1.3: 9 to 10 of 10 on the hard suite | 100% (30/30 tasks) | Not yet run | Not yet run | Not yet run |
| Conversation data | Paid plans (Lite and Pro): EBITDAI stores no prompts or workbook content. Free plan: conversations are logged and reviewed by our team. Meta Muse Spark 1.3 on Meta's Contributor tier, the only model on the free plan and on Lite, sends what you send it to Meta, which trains on it; the standard Meta tier, available on Pro, is not used for training | Used for Anthropic training unless you opt out | Used for OpenAI training unless you opt out | Free tier used for training; paid plans not trained | Not used to train foundation models |
| Proprietary modeling layer | All plans include EBITDAI's proprietary Modeling Layer | No: general agent | No: general agent | Spreadsheet-focused agent, own app | No: general assistant |
Competitor pricing, token rates, and limits from each vendor's public pricing, documentation, and user reports as of August 2026; details on the linked comparison pages. Subscription-implied token rates are estimates; assumptions are spelled out in our published benchmark. †Microsoft does not publish Copilot's per-token economics; the range assumes its Microsoft-hosted OpenAI models at OpenAI's public API list rates. Subject to change on their side.
EBITDAI never stores your prompts or workbook content on any paid plan. On the free plan, conversations are logged and read by our team to improve the product. The exact data path (what passes through our relay, what goes where, per plan and per provider) is documented in full on the security page.
Want to run on your own AI account instead? EBITDAI Pro runs EBITDAI on your own Kimi, DeepSeek or Gemini API key, and a $19/month Kimi Code subscription slots in the same way. Free and Lite run on Meta Muse Spark 1.3 Contributor, so they do not take your own key.
See how your data is handledInstall the add-in and sign up. The pane lives beside your workbook. You can find it here or by clicking the Add-ins button on the Home ribbon of Excel and searching “EBITDAI”.
The free plan gives you 150 AI requests a week on Meta Muse Spark 1.3, no card, and Meta trains on everything you send with that model. Pick a plan when you want more. Prefer to run on your own Kimi, DeepSeek or Gemini key? EBITDAI Pro connects it in one paste.
“Clean up this sheet.” “Explain what's driving the total in column F.” “Build me a 12-month budget.” The agent does the work in your workbook, every change visible.
All subscriptions include EBITDAI's proprietary modeling module.
Real session, one prompt: a complete 12-month budget with assumptions, revenue, costs, profit, formatting, and a chart.
Every balance rolls forward, every expense carries the right sign, statements tie. Built the way a banker would build it.
Assumptions in blue, formulas everywhere else, never hardcoded numbers. Bold headers, thousands separators, auditable structure throughout.
Debt schedules, working capital, capex and depreciation, SaaS cohorts, distribution waterfalls: house recipes, not improvisation. The methodology behind $1B+ of raised capital.
$5/month on Lite or $15/month on Pro, which runs Kimi, DeepSeek, Gemini and Meta Muse. Begin for free today; cancel anytime.
Beyond the add-in
The add-in is the self-service product. For finance teams, we build the software they are missing on the same platform, inside the workbook they already trust: every system connected, every monthly process automated, run by the agent in plain English.