EBITDAI puts an agentic AI analyst inside your workbook, with AI tokens included: install, sign in, and start modeling. Prefer to pick the brain? Connect your own Kimi, DeepSeek or Gemini key and switch models mid-session.
Free plan: 150 AI requests a week on Meta Muse Spark 1.3, and Meta trains on everything you send with it. No credit card.
Every AI lab now ships an Excel add-in: Claude for Excel, ChatGPT for Excel, Copilot, and Grok, each locked to its own models, priced at its own subscription, capped at its own limits. EBITDAI is the only agentic Excel add-in where you choose the model and pay the provider's actual price for it.
| Option | Monthly cost | How much AI you get | Model choice? |
|---|---|---|---|
| EBITDAI Lite | $5 all-in | 100x more usage than free, on Meta Muse Spark 1.3 Contributor, with no weekly request cap | Meta Muse Spark 1.3 Contributor only, and Meta trains on everything you send with it |
| EBITDAI Pro | $15 all-in | 7.5x more usage than lite across every model, including Kimi k3, the model that matched Claude in our benchmark | Kimi k3, DeepSeek V4.1 Flash, DeepSeek V4 Pro and Gemini 3.8 Flash, plus Meta Muse Spark 1.3 in both tiers, switch mid-session, or your own Kimi, DeepSeek or Gemini key |
| EBITDAI + Kimi Code | $34 all-in ($15 Pro + $19 Kimi Code) | Flat subscription: heavy daily use, no per-token bills. The only Excel add-in that can run on a Kimi Code subscription | Your own Kimi, DeepSeek or Gemini key, switch models mid-session |
| ChatGPT for Excel | $20 Plus for real limits | Usage caps by tier that reset on a timer | No: GPT models only, no own-key option |
| Claude for Excel | $20 Pro minimum, $100+ Max for heavy use | Usage limits measured in messages and hours | No: Claude models only, no own-key option |
| Microsoft 365 Copilot | $30/seat + M365 | Enterprise licensing, annual commitment | No: Microsoft picks the models, no own-key option |
Provider prices as of August 2026. DeepSeek cost estimate measured on EBITDAI's own benchmark model builds; your usage may vary. All AI accounts are purchased directly from the providers: EBITDAI never marks up tokens.
Point it at a broken column or an inherited workbook and ask what's going on. It reads the actual cells, not a screenshot.
Three-statement models, budgets, debt schedules, SaaS metrics: CFO-grade structure is built into its instructions.
Deduplicate, restructure, apply consistent formatting, split and combine sheets: the boring hour of your day, done in a minute.
On paid plans we never store your workbook contents or prompts: included-usage calls pass through our relay, which logs only token counts for billing. On your own key the path depends on the provider: DeepSeek and pay-as-you-go Moonshot calls go direct from your machine, while Kimi Code and Gemini cannot be called from a browser, so those pass through the same stateless relay, which stores nothing. Free-plan conversations are logged and reviewed to improve the product. One exception on any plan: pick Meta Muse Spark 1.3 on Meta's Contributor tier, the option shown as "trains on your data", and your prompts, replies and workbook values go to Meta, which trains on them. Free and Lite run that tier only; the standard Meta tier on Pro is not used by Meta for training.
One click from Microsoft AppSource. Works in Excel on the web and desktop.
Pick your plan and start modeling. Prefer to run on your own key? EBITDAI Pro connects your own Kimi (subscription or metered), DeepSeek or Gemini key.
"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.
EBITDAI Pro ($15/month) includes Kimi k3, DeepSeek V4.1 Flash, DeepSeek V4 Pro and Gemini 3.8 Flash on one shared allowance. Pro can also pick Meta Muse Spark 1.3 in two tiers: Meta's standard tier, which Meta does not use for training, and Meta's Contributor tier, where Meta trains on everything you send with it. Free and Lite ($5/month) run the Contributor tier only, which means Meta trains on everything you send with that model. On Pro you can connect your own Kimi key (flat-rate Kimi Code or metered Moonshot), DeepSeek key or Gemini key and switch models mid-session. Claude for Excel, ChatGPT for Excel, Copilot, and Grok's add-in are each locked to their own vendor's models; EBITDAI is the only agentic Excel add-in with model choice.
Because we run on DeepSeek, Kimi, Gemini and Meta Muse, whose token prices are a fraction of the big labs', the $5/month Lite plan carries 100x more usage than free on Meta Muse Spark 1.3 Contributor and the $15/month Pro plan carries 7.5x more usage than lite across every included model. If you would rather run on your own provider account, Pro does that, billed by that provider. On your own key, DeepSeek and pay-as-you-go Kimi calls go straight from the add-in to the provider; Kimi Code and Gemini cannot be called from a browser, so those pass through our stateless relay, which stores nothing.
On Pro, 7.5x more usage than lite across every included model, and unlike Claude's, your usage is dedicated to Excel, not shared with chat and coding. On Lite, 100x more usage than free on Meta Muse Spark 1.3 Contributor with no weekly request cap. Your remaining usage shows right in the pane; run out and you can buy extra usage on the spot. On Pro, running on your own key, you get as much as your own provider allows: DeepSeek is pennies per model, Kimi Code is flat-rate.
Yes. EBITDAI is currently the only Excel add-in that runs on a flat-rate Kimi Code subscription: $19/month for unlimited daily modeling with zero per-token charges. The plan itself is a bargain: third-party plan trackers show even Kimi's tier below Moderato allowing 300 to 1,200 calls per 5-hour window, against Claude Pro's reported 45 messages per window, and Kimi's tiers scale to 5x, 15x, and 30x base credits while Claude's upgrades are Max at $100/month (5x Pro) and $200/month (20x Pro). Those are each vendor's own ladder. We do not publish a multiple for EBITDAI against Claude or ChatGPT: the plans state their included usage as a multiple of the plan below, 100x more usage than free on Lite and 7.5x more usage than lite on Pro.
No. EBITDAI is strongest at financial models, that doctrine is built in, but the agent handles general Excel work: writing and fixing formulas, formatting, cleaning data, restructuring sheets, and explaining what a workbook is doing.
On paid plans, we never store it. Included usage on every paid plan passes through a stateless EBITDAI relay, which forwards the request and records only token counts for billing. On your own key the path depends on the provider: your own DeepSeek key and your own pay-as-you-go Moonshot key are called directly from your machine, so the key and the workbook never touch our infrastructure, while a Kimi Code key and a Gemini key cannot be called from a browser, so those requests pass through the same stateless relay, which keeps nothing: no prompts, no workbook content, no keys. The free plan is the exception: free-plan conversations are logged and read by our team to improve the product, and deleted after 90 days. Full detail at /security.
From $5/month. Any model, your key, at cost. Free to try.