State of M&A Data Rooms — Q2 2026 Read the report →

Automate Peony with an AI Agent (Claude Code or Codex)

Let Claude Code or Codex build data rooms, set group permissions and read analytics in Peony for you, by driving a signed-in ego lite browser. No API key, any plan.

Last updated October 23, 2026

You can hand repetitive admin work in Peony to an AI coding agent such as Claude Code or Codex. The agent opens Peony in ego lite, a free browser built for AI agents, and clicks through the same screens you would: it creates rooms and folders, uploads files, sets up groups and reads analytics. It needs no API key, because it works inside your own signed-in session.

Who can do this: any admin (Owner or Member), on any plan Where: a Mac, with ego lite and Claude Code or Codex installed

Set It Up

  1. Download ego lite from lite.ego.app and finish its onboarding. Onboarding installs the ego-browser command and adds the ego-browser skill to Claude Code and Codex.
  2. In ego lite, open app.peony.ink and sign in with the code Peony emails you. The agent uses this session, so you only sign in once.
  3. Open Claude Code or Codex in a terminal and describe the job in plain English. Name the data room and team, and say what the agent must not change.
  4. When the agent reports back, open the room in Peony and check the result before you share anything.

Prompts We Tested

We ran each of these jobs end to end with Claude Code and ego lite against demo rooms in Peony. Copy one and change the names.

  • Build a room from a folder on your computer. "Create a data room called Project Juniper in the team 01. Advisor Team and recreate the folder ~/Deals/Project Juniper inside it, subfolders included. When you're done, compare the number of files in Peony with the local folder and upload anything missing."
  • Set up a group with limited access. "In Project Juniper, create a group named First-round bidders with Link access. Give it View on 01. Process Letter & Teaser and 02. Confidential Information Memorandum, No access on every other folder, and turn Screenshot protection to Basic. Don't invite anyone. Take a screenshot of the permission tree when you're done."
  • Audit what is set up, without changing it. "Open Project Juniper. List each top-level folder with its file count, then tell me which folders First-round bidders can view and which security controls are on. Do not change anything."
  • Summarize engagement. "From the Analytics tab of Project Falcon, list the five most active visitors by views and by time spent, and the five most viewed files."

What to Expect

  • Everything is done as you. The agent signs in with your session, so every room, folder and group it creates belongs to your account.
  • It takes minutes, not seconds. Our read-only check of a 23-file room took between 3 and 7 minutes. The agent reads each screen before it acts.
  • It recreates folders one by one. The agent can't operate your computer's folder picker, so instead of Folder upload it creates each folder with New folder and uploads its files with File upload. The result is the same tree.
  • It reports back in the chat. Ask for a table, a count or a screenshot and you get it in the terminal.

Notes and Limits

Tell the agent what it may not do. Deleting a file, folder or data room in Peony is permanent, and an Email access group sends invite emails as soon as it is saved. Say "don't delete anything" and "don't invite anyone" in every prompt that touches a live room.

  • Check counts after an upload. In our test the first pass missed one of 23 files. Asking the agent to compare Peony with the local folder caught it.
  • Check permissions after a group is created. A new group starts with View on every folder. Tell the agent which folders should be No access, and look at the permission tree yourself before you send the link.
  • What the agent reads goes to its AI provider. File names, visitor emails and analytics on screen are sent to Anthropic or OpenAI as part of the conversation. Check your firm's AI policy before you point an agent at a live deal room.
  • Audit trail on Enterprise. Admin changes can be audit-trailed on Enterprise, so the agent's changes are recorded under your account. To log the AI itself as its own actor, use the Enterprise Connect your own AI setup.
  • Not a replacement for the API. For jobs that run on a schedule or from your own servers, such as syncing a CRM, use API access, which is a Deal Team feature enabled on request and included on Enterprise.
  • Mac only for now. ego lite runs on macOS. It doesn't work on Windows yet.
  • ego lite is a third-party app. It is made by Citro Labs, not Peony. For install or browser problems, see lite.ego.app.

Common Questions

Do I need an API key or a particular plan?

No. The agent works through your own signed-in browser session, so it can do whatever your admin account can do on your current plan. Features that need a higher plan still need that plan.

Does the agent show up in the audit trail?

On Enterprise, yes: admin changes can be audit-trailed, and the agent acts as you, so its changes are recorded under your account. To log the AI as its own actor, use the Enterprise Connect your own AI setup. On any plan, the agent's chat transcript is also a record of what it changed.

Can I use a different agent browser?

Yes. Any tool that lets an agent drive a browser you are signed in to works the same way. We tested ego lite with Claude Code because it needs no setup beyond the install and inherits your logins.

Still stuck? Use the in-app chat or email sean@peony.ink.