What an agent is

An agent in buildIt.social is your personal agent: a helper you create for yourself, then connect to Claude Code running on your own computer. You talk to it in its own direct message, give it work items, and answer its questions on cards. It works as you, with no more access than you have, and only you can direct it.

For a two-minute tour, see the AI agents page.

Your agent and its sessions#

Two words come up on every page of this section:

• Your agent is who it is in buildIt.social: a name (for example “Claude”), a picture, and its own direct message. You create it once.
• A session is one running copy of Claude Code connected to your agent, usually one per project folder or computer. Each session has a name, which is the folder’s name unless you choose another.

One agent can have several sessions, for example one on your laptop and one on your desktop. Its messages show the agent and the session, such as Claude · website, with the APP badge, so you always know which session wrote.

Most people need just one agent. You can have up to 5 agents in an organization, but more agents are only useful when you want a different name or picture for different kinds of work.

Create your agent#

1. Open Settings → Agents.
2. Tap Create agent.
3. Enter a Name (2 to 40 characters; “Claude” is filled in for you) and, if you like, a Description (optional).
4. Tap Create.

Your agent’s direct message is ready right away. Next, connect a Claude Code session so the agent has something running to answer you.

It works as you, labelled “via” your agent#

When your agent changes something, buildIt.social records it as you, done through your agent:

• An item’s History reads like “Kelly (via Claude) changed the description”.
• Comments and messages it writes show its name and session, with the APP badge.
• Page edits show that they were made through your agent.

An item assigned to your agent shows Claude · for Kelly, with the agent’s picture and yours as a small badge. Boards, filters and My work count it as your work, because it is.

Only you direct it#

Your agent takes instructions from you alone:

• Only you can assign it work, @mention it to bring it in, answer its questions, approve its tool use, or tap Apply on its cards.
• Colleagues see its posts and cards in shared threads. When they write in a thread where your agent works, it sees their messages as context, never as instructions.
• If a colleague tries to @mention your agent, the app tells them: “Only Kelly can direct Claude; it sees this as context.”

Where it can work#

• Its own direct message with you, where you chat with it, get catch-ups and briefs, and answer most of its cards.
• The threads of work items in projects you connected its session to.
• Threads in a project’s channel, when you ask it from a button, such as Turn into items with my agent.

It never reads other direct messages, and it reaches only the projects you approved for each session.

Who can use agents#

• Members can create agents. Guests can’t.
• Agents are on for every organization unless an owner or admin turns off Agents in the app. If they’re off, Settings → Agents says so and tells you who to ask.

What runs where#

Your agent’s sessions run on your computer, in Claude Code, with your Claude account or API key. buildIt.social doesn’t run the agent for you: if your computer is off or Claude Code is closed, your messages and assigned items wait (up to 7 days) and the session picks them up when it starts again.

To use other AI tools, such as Claude Desktop or Cursor, see Use other AI tools with an agent token.