CATO Chat and Workspace
Ask for analysis, literature work, code, figures, and manuscripts in a conversation, and open the results beside it.
Use the files as the record, not the chat message. Say which data to use, follow the run in Chat, then check the files, figures, code, and citations in the Workspace.
Modes and Effort
| Mode | Use it for | What to expect |
|---|---|---|
| Standard | Analysis, coding, writing, and routine research | The default |
| Research | Harder methods, multi-step evidence work, and difficult analyses | Takes longer and uses more of your allowance |
Change modes with the model control in the composer; /standard switches back to Standard.
The reasoning-effort slider in the composer sets how much CATO reasons before it answers. Low is fastest, Medium suits most analyses, High reasons more on every step, and Max is slowest and uses the most of your allowance. Your choice applies to the message you send and stays as your default. Until you choose, Standard uses High and Research uses Max.
Say What to Use
CATO does best when it knows exactly which data and papers you mean. Use @ to mention a dataset, name the file or version in your message, or send selected papers from the Library.
- Give exact filenames, versions, variables, cohorts, and timepoints.
- Say whether CATO may use general knowledge or only the sources you selected.
- Remove attachments that do not matter to the question.
- In follow-up work, say which generated dataset or figure version you mean.
Write the Message
Send the message.
Insert a line break.
Edit your last message when the input is empty.
Browse commands such as mode changes and dataset profiling.
Close a menu without sending.
Ask CATO to plan first when the method, inclusion criteria, or model need your approval. Read the plan, then tell it to go ahead or what to change.
An Example Request
Objective: Estimate the adjusted treatment effect on the 12-week score.
Evidence: Use @analysis_v3.parquet and @data_dictionary.pdf.
Population: Include randomized participants with a baseline measurement; explain all exclusions.
Method: Fit the prespecified model, check assumptions, report effect size and 95% CI, and run one sensitivity analysis.
Outputs: Save a model table, diagnostics, analysis notebook, and editable publication figure.
When to stop: Stop and ask if variable mappings conflict with the dictionary.
Following a Run
While CATO works, the status line shows what it is doing, and long thinking shows a timer. For work with several steps, a run rail shows what CATO is doing now, and Plan on the rail opens the steps. CATO may pause to ask you a question; answer it in place to continue.
- Claims in the response carry small marks tying them to the evidence behind them. A claim CATO could not find in its work has a dotted underline; hover over it, tap it or select it to see why.
- A meter shows how much of the conversation's working context is in use.
The Project Computer
Each project has its own computer, where analyses run and files are kept. The Workspace panel shows its name, with Files, Shell, and Memory views. Files stay between conversations, and uploads land in the folder you are browsing. You can rename the computer. It starts when CATO needs it, which can take a few minutes.
Leaving and Coming Back
CATO keeps working if you leave the page, reload it, or lose your connection. A long command keeps running, and CATO carries on when it ends. Reopen the same conversation to see where it got to.
- Reconnecting means the page is catching up. Wait for it before sending the request again.
- Press Stop once to end a run, and give it a moment to finish stopping. To end one command, press Stop beside it in the Shell view.
- If a response stops early, what was finished stays in the conversation and in Files.
The Workspace
Files
Browse the project computer: folders, drag to move, upload, preview, rename, and see where a file came from.
Shell
See each command CATO runs, with its output, how long it took, and a Stop button while it runs.
Memory
Read, add, edit, and delete CATO's notes on this project.
Figures
Open plots and scientific viewers, edit SVG figures, comment on them, compare versions, and download them.
Notebooks
Read the code and its output, download the script or notebook, and compare versions.
Manuscripts
Edit Markdown or LaTeX, preview the result, and export it as Word or PDF.
Provenance
In a file's menu: the command that made a version, what it read, and the part of the conversation it came from.
Review
What Double-check this answer found about the answer behind a file.
The tabs you had open come back when you return to the conversation. If you have unsaved edits, CATO asks you to save or discard them before you switch versions or close the tab.
Figures and Scientific Viewers
Figures appear in the conversation and in the Workspace. Open an SVG figure in the editor to select, move, group, layer, and restyle elements and add text. 3D and other interactive views have their own controls.
- Saving an edit makes a new version, and the earlier ones stay. Any version can be edited, the original included, and the versions list says which one an edit came from (“v3 · from v1”).
- The default version is the one that opens first and that the chat shows and cites. A new version becomes the default. To go back to an older one, open it and choose Use this version. Public citation links already made keep their version.
- Use accessible colors, legible labels, units, uncertainty, and complete captions.
- After editing, check that the data marks and axes did not change.
- Before you submit, check the downloaded file and the journal's dimensions.
- A public figure link can be opened by anyone who has it, so do not use one for restricted data.
Notebooks and Code
CATO writes and runs Python or R in an isolated sandbox and saves the notebooks and files in the project. Code runs in CATO's session, not yours: ask for the change you want and it re-runs the analysis.
- Read the code, package versions, seeds, inputs, and output paths.
- Ask CATO to check that a package is installed before you plan around it.
- Re-run important results from a clean state before you publish them.
Manuscripts, Readers, and Comments
You can edit a manuscript in Markdown or LaTeX, preview it, and export it as Word or PDF. In PDF, HTML, and transcript readers, select a passage and choose Comment; on a figure, choose Comment and click a point. Comments go with your next message.
Check the work yourself. Check every quotation, citation, number, table, figure, and statistical conclusion against its source.
Memory
CATO keeps notes so that a later conversation can pick up where the last one stopped. In each turn it sees one line per note and opens a note when it is relevant. There is no switch to turn memory off, but you can read, edit, or delete any note.
- Project notes are in the project's Memory tab and are shared with everyone on the project. You can also add your own there.
- Your own notes, about how you work, are in Settings → Memory. They follow you across your projects and only you can see them.
- A project note may record a result and the file it came from. CATO checks the file before relying on the note, and what you say in the conversation takes priority.
- If CATO seems to be using something outdated, look for the note and edit or delete it.
Sending Papers to Chat
Select papers in the Library and choose Send to Chat, then pick Ask with context, Analyze and extract, Draft with citations, or Plan next experiments. Papers CATO can fetch in full text also open in the Workspace, where you can read and comment on them.
Open the cited paper and confirm that it supports the claim. Learn more in the Library and Deep Search guide.
Double-check This Answer
The button sits under CATO's latest answer. A second model rereads the answer against the record of what CATO ran and read, and lists any claim the record does not support. It does not rerun the analysis or judge whether the method was right, and it is not peer review. You can add your own criteria in Settings → Specialists.
When Your Usage Runs Out
Standard and Pro include a monthly usage allowance. When it is used up, new messages are refused until it resets, and CATO tells you the date. A response already being written finishes, and a run with more to do stops there. Pay-as-you-go has no monthly cap. Settings → Billing shows how much you have used.
A Checklist for Your Request
- Is the research question specific, and the population clear?
- Have you named the right files and versions?
- Are variable definitions, units, missing-value codes, and exclusions clear?
- Did you ask for diagnostics, uncertainty, sensitivity checks, and limitations?
- Did you say which outputs to save?
- Will you check the code, results, figures, and citations before you use them?
For examples, continue to Best Practices or use the Quick Reference.