Skip to main content

Documentation Index

Fetch the complete documentation index at: https://university.gumloop.com/llms.txt

Use this file to discover all available pages before exploring further.

VideosAgent Inboxes

You can now email any Gumloop agent so they can start handling tasks for you. Just loop them in and they can finish the job.

You could use this to triage inbound leads, manage vendor proposals, or answer security questionnaires. In the demo, the agent already understands Gumloop’s security policies through a skill, so it can fill one out without me doing it manually.

To make that possible, the first step is to give the agent an inbox. That creates an @gumloopagents.com address for the agent.

Any email sent to that address from someone on your team gets handled by the agent. In the walkthrough, the security questionnaire is forwarded to the agent’s address.

Once the email lands, the agent receives the thread and starts working on it as if the whole exchange had been copied and pasted directly into Gumloop.

In this example, the agent reads the Google Sheet attached by the customer and starts answering the questionnaire automatically.

When it’s done, the agent replies back into the same email thread. The outcome is simple: no more answering security questionnaires by hand, and no more manually triaging repetitive vendor requests. You just loop in your Gumloop agent.

Agent Inboxes

Agent Inboxes give your agent an email address, so your team can send work to an agent straight from email.

Some work starts in Gumloop. A lot of work still starts in email.

Agent Inboxes bridge that gap. You give an agent its own email address, then forward a thread, CC the agent, or loop it in the same way you would loop in a teammate.

What problem this solves

Email is where a lot of messy, real work shows up. A vendor sends a questionnaire. A customer asks for a spreadsheet to be filled out. A lead asks a question that needs context from three different tools.

Without Agent Inboxes, someone has to copy that context into Gumloop by hand. With Agent Inboxes, they can forward the thread, CC the agent, or loop it in and keep working from the inbox they already use.

How it works

  1. You give an agent an inbox.
  2. Gumloop creates an @gumloopagents.com address for that agent.
  3. Your team forwards an email, CCs the agent, or loops it into a thread.
  4. Gumloop checks whether the sender is allowed to use that agent and its tools.
  5. The agent reads the email, attachments, and thread context.
  6. The agent completes the task and replies back into the thread.

The agent works from the full thread context, not a stripped-down prompt. That means it can see the same message history and attachments a human teammate would have received.

How permissions work

Agent Inboxes follow the same permission model as chatting with an agent in Gumloop. Forwarding an email does not give someone extra access.

When an email comes in, Gumloop checks two things:

  • Agent access: the person forwarding the email must have access to the agent.
  • Tool access: the agent can only use the underlying tools and data that person is allowed to use.

In other words, emailing an agent works like asking that same agent for help in the web app. If the sender could not use the agent or a connected tool there, the inbox does not bypass that.

What to remember

Agent Inboxes are not just email notifications for agents. They let an agent participate in the place where work already happens.

That matters because it lowers the effort for the rest of your team. They do not need to learn a new interface to get value. They can just loop in the agent.