Import prompts in bulk
There is no CSV upload. The three ways to add many prompts quickly are discovery multi-select, the API, and the MCP connector.
There is no CSV import. That is worth stating plainly, because export exists and people reasonably assume the reverse does too. Here are the three ways to add prompts at volume.
What you'll learn
- The three bulk paths and when each is right
- How to go from a spreadsheet to a populated workspace
- What to check after any bulk add
Option 1: Accept a batch from discovery
Best for a new client where you do not yet have a prompt list of your own.
- Open Prompts in the client's workspace.
- Open discovery. It proposes prompts from the site, the category and the answers already read.
- Select the ones a buyer would actually ask, and add them to a topic in one action.
Review rather than accept wholesale. A proposed prompt that reads like a search phrase will produce an answer that names nobody.
Option 2: The public API
Best when you already have the list in a spreadsheet, and best by a distance when you are setting up several workspaces.
The endpoint takes one prompt per call, with an optional topic and an optional platform list, so a bulk load is a short loop over your rows. A twenty-line script will populate a workspace faster than any interface.
- Create a key in API settings.
- See API access for authentication and the endpoint.
- Duplicate prompts are rejected rather than silently merged, so a re-run of a partially completed load is safe.
- Unknown platform ids are rejected rather than filtered away, which means you will hear about a typo instead of creating a prompt that is checked on nothing.
Option 3: The MCP connector
Best if you already work in an AI client such as Claude or Cursor. The connector exposes prompt-adding as a tool, so you can hand it a list in conversation and have it create them in the right workspace.
See MCP.
Going from a spreadsheet to a workspace
- Clean the list first. One question per row, phrased as a question, no client name in the generic ones.
- Decide the topics before you load. Topics are how the lens groups everything and how reporting rolls up, and setting them afterwards is tedious.
- Load into one workspace and look at it before you run the same load across ten clients.
After any bulk add
- Check the count against your plan. Prompts consume coverage on every surface every cycle. See coverage and credits.
- Check the intent mix. A bulk load usually skews informational, because those are the easiest questions to write.
- Wait a cycle before judging anything. A fresh prompt with no answers yet is not a prompt you are losing.
Exporting
Export to CSV does exist, per prompt selection, which is the fastest way to get a client's prompt set into a spreadsheet for review in a kickoff.
Try this in Outercite
Export one client's prompt set to CSV, sort it by whether you could imagine a buyer saying it out loud, and delete the bottom half. Then use the remainder as the template for your next client.
Related
Add prompts
The single-prompt path and the topic conventions.
API access
Keys, authentication and the endpoints.
MCP
Working with your workspace from an AI client.
Last updated on
