Key Concepts
Tool: An app the agent can use. You decide what it is allowed to do in each one.
Reading tool: Lets the agent look things up, like a web page or your analytics. The agent only takes information in, so there is little that can go wrong.
Writing tool: Lets the agent save or send something, like a doc or a message. These change things outside Profound, so they deserve more care.
Quick Reference
Permission settings (one per tool)
- Ask: the agent checks with you first
- Allow: the agent can go ahead
- Block: the agent cannot use the tool at all
Good starting point
- Allow reading
- Ask before writing
- Block anything the agent should never touch
Add a tool
- In the agent builder, open the Tools section and click Add tool
- Search the list (options include Firecrawl, Exa, Slack, Google Analytics, and Microsoft 365) and pick tools your team already uses
- Click Connect account, or use the default account if one is already set
- Open the Integrations page and set Ask, Allow, or Block for each tool
Example setup: content brief agent
- Reading: Firecrawl, set to Allow, so the agent can study the pages answer engines already cite
- Writing: Google Docs or Slack, set to Ask, so the brief can be saved or shared, with a check-in before anything is written
Whose account the agent uses
- By default, a tool runs on the account of the person running the agent
- When a teammate runs your agent, they need the tool connected on their side too
- On a published agent’s page, the Tools tab shows what anyone running it needs
Notes
- Profound data is available to the agent without connecting anything, so it can read your visibility and citation data from the start.
- The two jobs to plan for are reading (get the inputs) and writing (land the output where your team will find it).