Start a new workbook
create_workbookMakes a new workbook from EXACTLY ONE of: instruction — one sentence the way a person would say it ('find 30 US importers of loose leaf tea and the buyer at each'), which starts a run that researches and writes it, is billed to this account and takes minutes; document — markdown you have written, which becomes version 1 at once; or sheets — tabs, each {name, markdown}, which become version 1 at once. Writing content yourself runs nothing and costs nothing. Nothing that already exists is changed.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| mode | No | What this workbook's runs are, from now on. `harnessroom` (a new workbook's default): the instruction is a request, and a harness researches and writes the document. `latticeroom`: the instruction is a draft someone already wrote, whole — markdown, its links kept — which comes back exactly as written, with the run's inference beside it (read_workbook's `inference`: each claim cited, self-reported, secondary only, contradicted or with no source, and its sources); each entry under its References or Bibliography heading is looked up in Crossref and on the web and marked found, found with other authors or year, or not found, and a sentence citing it — [3], (Smith et al., 2020) — is read against the work found; attach data files (.json, .csv, .tsv) to settle the draft's own figures. `graphroom`: runs perform the procedure drawn for the workbook in the app, and refuse until one is drawn. Omit to keep the workbook's current mode. | |
| name | No | Its name. A name another workbook has is told apart by this one's id. | |
| model | No | The model its runs think with, as configure_workbook names them — set before the first run. | |
| folder | No | A folder to put it in, e.g. 'Outreach'. Made if missing. Omit for the root. | |
| sheets | No | The first version as tabs, in order. | |
| context | No | Explain in 15-25 words, in third person, why this tool is called and how it supports the user's goal. For analytics only. You MUST describe only the abstract purpose of the tool call. NEVER include, repeat, paraphrase, or infer personal, sensitive, or identifying information from the user request or tool results, including names, emails, phone numbers, IPs, IDs, or credentials. You MUST generalize specific entities into roles such as "a user", "the customer", or "an account". Example: "Retrieving a customer's recent orders to investigate a billing issue and help support determine the appropriate resolution." | |
| document | No | The first version as markdown, written by you. | |
| llm_model | No | The exact model identifier you (the assistant) are running as, taken from your system prompt or environment (e.g. "claude-opus-4-8", "gpt-5.2"). Used for analytics only. If you do not know your model identifier with certainty, pass "unknown" — never guess. | |
| attachments | No | Files to attach to the workbook BEFORE the run starts, so it reads and cites them — each {name, content_base64, content_type?}, 32 MB at most. They stay attached for every later run; list_attachments shows them. | |
| instruction | No | What the workbook should be, in one sentence. A billed run writes it. | |
| time_limit_seconds | No | How long its first run may take, for a run a person waits on: stopped there, it delivers the draft it has. Set before the run starts, so one call makes and starts it. |