Calculators & scenarios
Estimate effort, compare assumptions and keep a report of the result. The included project-estimate package demonstrates this.
Give the workflow a tool of its own.
Workbooks turns recurring spreadsheet jobs into small operational tools. Your workspace agent builds the logic. Your team gets a clear interface. Your company owns both.
Self-hosted on Cloudflare. Public source release in preparation.
Change the assumption. See the result.
Cost = days × cost per day.
Margin = (price − cost) ÷ price × 100.
Illustrative values in one consistent currency. Overheads and taxes are excluded.
Try it here. This illustration does not save data.
The Workbooks concept, in under a minute.
It starts as a spreadsheet. Then the spreadsheet becomes a process. What can you edit? Which formula can you trust?
Workbooks gives that process a tool of its own. Describe the job to the agent in your company workspace.
The agent builds code you can review. Your team changes an assumption and sees the result. Ten days. Eight days. A different margin.
Save the record. Open the report. Run recurring checks on a schedule. Ask the agent when the workflow needs to change.
The code lives in your repository. The data lives in your Cloudflare account. Your company owns the tool.
Keep spreadsheets for exploration. Give recurring work a tool of its own. Spark Workbooks.
The conversation happens in your company workspace, where your agent already has the relevant context. Workbooks does not add another chat window.
“Build an inspection tracker. Show what is due and check it every morning.”
Your agent creates the workbook package: data, logic, actions and interface.
Check the assumptions and outputs with the person who knows the work. Changes to the code stay reviewable. Data migrations have an explicit plan.
People use the browser. Records persist. Scheduled actions run without an open tab. New requirements go back to the agent.
Spreadsheets are excellent for quick exploration and open-ended analysis. Workbooks earns its place when the same process needs a dependable interface, tested logic and recurring actions.
| What matters | A spreadsheet | A workbook |
|---|---|---|
| The interface | A flexible grid. People need to know which cells to edit. | A form, calculator or report built around the job. |
| The logic | Formulas and scripts sit alongside the sheet's data. | Definitions and migrations live in Git, separate from live records. |
| A change request | Someone edits the sheet, formulas or connected scripts. | Your workspace agent changes the source, tests it and deploys the reviewed change. |
| Recurring work | Possible through macros, scripts and automation services. | Named actions and scheduled checks are part of the workbook package. |
| The result | A sheet to interpret, export or turn into a report. | A purpose-built view or saved-record report for the person using it. |
| Ownership | Depends on your spreadsheet and hosting setup. | Your code repository, Cloudflare account and live data. |
Estimate effort, compare assumptions and keep a report of the result. The included project-estimate package demonstrates this.
Store dates, calculate what needs attention and run a daily check. The inspections example records signals and quiet runs.
Give the team a clear view over required reference data, with the definitions kept beside the code.
One company installation. A separate SQLite database for each workbook. Definitions and logic in Git. Your own Cloudflare account and hostname.
One building block of Spark OS, usable with other agentic workspaces too. Use your ERP or CRM as the system of record. Add a workbook where a specific job needs its own tool.
Ownership also means operation: someone must manage access, deployments, backups and updates. Spark can build and install the tool, then hand over its source and operating instructions.
No. The agent needs to understand its formulas, assumptions and workflow, then build and test an equivalent tool. There is no universal spreadsheet converter.
The self-hosted runtime is built and MIT-licensed. The public repository release is still in preparation. Contact Spark to discuss a pilot; this page does not promise instant public installation.
This version runs in your Cloudflare account using Workers and Durable Objects. It is not a Docker application for an arbitrary server.
People use the workbook according to their access role. Changes to code happen through your development workflow. Your workspace agent helps author them; testing and review still matter.
Start with the spreadsheet your team works around.