Getting Started
Create a project, connect evidence, run a reproducible analysis, and inspect every resulting artifact.
The core mental model: a project holds your source data, literature, conversations, memory, and versioned outputs. Chat coordinates the work; the Workspace is where you inspect and refine it.
1. Create a Project
- From the projects screen, select New Project.
- Give the project a clear study, client, or research-question name.
- Add a description, labels, and the citation style you expect to use.
- Open the project. The guided tour can introduce the current interface and can be replayed from Help.
Keep unrelated studies in separate projects. Project-scoped memory, literature context, and artifacts are designed to remain inside that research boundary.
2. Add Research Data
Open Database, select Upload, or pressCtrl/Cmd+U. You can also connect a project folder from Google Drive.
Common formats
CSV, TSV, Excel, Parquet, SAS, SPSS, Stata, JSON, PDF, images, HDF5, H5AD, FASTA/FASTQ, VCF, BED, PDB/CIF, DICOM, NIfTI, mzML, MATLAB, NumPy, and archives.
Before analysis
Check the detected type, row and column information, missingness, and any upload warning. Descriptive names such as baseline_labs.csv produce clearer prompts than generic filenames.
Upload limits are 100 MB on Standard, 500 MB on Enterprise, and 1 GB on Pay as you go. Compressed archives are additionally protected by archive and decompressed-size safety limits. See Data Management for the full guide.
3. Learn the Project Workspace
Database
Source datasets, folders, generated files, metadata, and external data connections.
Chat
Conversations, analysis modes, Plan mode, review controls, and durable background runs.
Literature
Deep Search and Library for search, evidence extraction, watches, and synthesis.
Workspace
Files, figures, notebooks, manuscripts, annotations, provenance, logs, and version history.
Use Ctrl/Cmd+Kfor the command palette, Ctrl/Cmd+1for Database, and Ctrl/Cmd+2for Chat.
4. Choose an Analysis Mode
| Mode | Best for | Behavior |
|---|---|---|
| Standard | Everyday analysis, writing, data exploration, and balanced research tasks. | The default balance of quality, latency, and AI usage. |
| Fast | Quick questions, iterative edits, summaries, and lightweight exploration. | Optimized for lower latency and billed at the Fast-mode usage multiplier shown in the product. |
| Research | Complex methodology, multi-stage analysis, evidence synthesis, and maximum-depth reasoning. | Allow more time for extended planning, tool use, and verification. |
Use the mode control or the /standard, /fast, and /researchcommands. Mode names describe CATO behavior; underlying providers may change without altering your workflow.
5. Ask a Grounded Question
Attach or select the exact evidence CATO should use. Type @ to mention a dataset and / to browse available commands. Press Shift+Tab to request a reviewable plan before execution.
“Profile @baseline_labs.csv. Report missingness and outliers, but do not impute yet.”
“Compare the primary outcome by treatment arm with effect size, 95% CI, assumption checks, and a publication-ready figure.”
“Using only the selected papers, build an evidence table and identify contradictions that need manual review.”
A strong prompt states the question, relevant variables, population or subset, desired method, required checks, and output format. If a definition is ambiguous, ask CATO to confirm it before running code.
6. Follow Long-Running Work
CATO persists active response streams and background sandbox jobs. You can move to another conversation or reload the page and reconnect to the same run. Background job cards show progress, completion, artifacts, failures, and cancellation controls.
- Use Cancel when the question or data selection is wrong.
- Do not submit the same request repeatedly while a run is reconnecting.
- If generation stops, reopen the conversation first; durable output may already be recoverable.
- Billing and usage limits can interrupt work mid-run; the interface preserves the useful partial state when possible.
7. Inspect and Refine Artifacts
Open generated files and figures in the Workspace instead of treating the chat message as the final record. CATO stores artifact provenance, versions, execution context, and the conversation that produced the output.
Files and data
Preview, edit supported text or code, choose a version, and reuse generated datasets.
Figures
Inspect interactive views, edit supported SVG figures, annotate, and export.
Notebooks and manuscripts
Continue analysis in persistent kernels or refine Markdown/LaTeX with preview and export.
If an editor reports unsaved changes, save or discard them before switching versions or closing the artifact. Reviewer findings and annotations are guidance—not a replacement for validating data, code, statistics, and citations.
8. Search and Curate Literature
Use Literature → Deep Search for parallel scholarly retrieval, then save relevant works to the Library. CATO searches 18 open scholarly sources, deduplicates records, can retrieve available full text, and supports evidence extraction, networks, watches, and document generation.
Fetched papers can open in the Workspace and selected literature can be sent to Chat as a bounded context packet. Always verify bibliographic details and source-grounded claims before publication. See the Literature Intelligence guide.
9. Control Memory
Memory can retain stable preferences and facts grounded in your messages. Project memory stays with one project; global memory can apply across projects. Review memory settings to enable or disable recall, automatic learning, and cross-project use, and edit, pin, or delete individual entries.
Do not use memory as evidence. Generated findings are not automatically promoted to research facts. Keep scientific conclusions in cited artifacts, datasets, and manuscripts.
Plans, Uploads, and AI Credit
| Plan | Price | Included AI credit | Max file |
|---|---|---|---|
| Standard | $20/month | $18/month | 100 MB |
| Enterprise | $100/month | $80/month | 500 MB |
| Pay as you go | Metered | Usage billed as incurred | 1 GB |
Billing in Settings is the source of truth for your account, taxes, current prices, usage, renewal date, and any institutional or negotiated terms. AI credit is a monetary usage allowance, not a fixed message count.