Build your first research team

Choose specialists, review their limits, and give one clear setup file to your AI assistant.

Start with the example Open the builder
A researcher receives separate advice from clinical, statistical, literature, writing, and peer-review specialists, makes the decision, then approves one path to execution.

Why use a team instead of one agent?

A single AI conversation can mix evidence review, scientific judgment, editing, and execution. That makes it hard to see which kind of advice you are receiving.

ResearchCord gives each specialist a defined job and explicit limits. You can compare their advice, resolve disagreement, and decide what happens next.

The authority stays with you.

Specialists recommend. You decide. Your assistant acts only after the decision is clear.

Choose the closest starting point

You do not need to design a team from scratch. Pick the project type that is closest to your work, then change only what is wrong.

Example: For a cardiac rehabilitation manuscript, choose Write or finish a paper.

Review who is on the team

A useful team is not the largest team. Keep the specialists who have a distinct job in this project.

PostdocCoordinates the work and keeps requests moving.
Clinical expertChecks clinical meaning and interpretation.
BiostatisticianReviews design, analysis, and statistical claims.
Citation auditorChecks whether sources support each claim.
Medical writerImproves clarity without changing the science.
Peer colleagueChallenges assumptions before outside review.
Why this matters: Role limits keep statistical advice from becoming clinical judgment, and editing from becoming scientific approval.

Preview the team, then create the handoff

Use Preview team file to check the roles and limits. When the team looks right, choose the handoff that fits how you work.

Copy a sharing link

Send the link to your AI assistant or collaborator. The link contains the team configuration and setup instructions.

Download the team file

Save SKILL.md. In Claude Code, place it at ~/.claude/skills/project-name/SKILL.md. In another assistant, attach or point to the file.

Check before sharing: The file should name the right specialists, preserve your authority, and state that execution waits for approval.

Ask one question with clear boundaries

Name the specialists you want, the decision you need to make, and what must wait until you approve it.

Cardiac rehabilitation manuscript Example consultation request 
Ask the Biostatistician and Clinical Expert whether our primary outcome analysis supports the conclusion.

Return their advice separately. Flag any disagreement or uncertainty. Do not edit the manuscript until I decide what to accept.
What ResearchCord adds: Each specialist stays in scope, recommendations remain separate, and disagreement becomes visible instead of being blended into one answer.

Compare the advice and make the call

Read the recommendations, ask follow-up questions, and record what you accept, reject, or modify. Only then ask the appropriate role to execute the decision.

A plain-language ResearchCord correspondence record with the request, response, and execution update in separate sections.

What the record gives you

ResearchCord correspondence keeps the reasoning visible in a browser-readable file.

  • The question you asked
  • Advice from the responsible specialist
  • Your decision or requested change
  • The later execution update
Research safety: This supports structured work. It does not replace scientific review, clinical judgment, ethics review, or regulatory approval.

Ready to try the example?

Open the builder, choose the paper starting point, and review the suggested team before changing anything.

Open the builder