Recommended flow
The cleanest first run is narrow: ask one question, ground it with files or sources, then turn the useful result into an artifact.
Step 1
Start a research session
Open the agent, ask a narrow first question, and attach any source files that should shape the answer.
Step 2
Ask for evidence, not just prose
Cortexa can search live tools and cite retrieved sources. Ask for mechanisms, evidence tables, trade-offs, or open questions.
Step 3
Use experiments for analysis
Run Python cells when a question needs data cleaning, statistics, charts, or model-ready transformations.
Step 4
Package the result
Generate a brief or presentation once the session has enough context. The artifacts stay tied to the source conversation.
A good first request
“Summarize the current evidence for X, cite primary sources, separate established findings from hypotheses, and list the next three analyses I should run.”
Prompt patterns
Good prompts tell the agent what evidence matters, which files to inspect, and what shape the answer should take.
Ask for a structured answer with citations.
Name the dataset or file you want the agent to use.
Tell the agent whether you want a fast scan or a deeper investigation.
Use follow-ups to refine, critique, or convert the answer into an artifact.