Agents in your Workflow

An Agent connected as you works under your name. An Agent can also hold a seat of its own: an account it signs in with, tasks assigned to it, and time recorded under its own identity, stamped as agent work in the ledger. In this guide you set one up end to end: create its account, connect it, hand it a task, watch the work come back through the board, and put it on a lane so it keeps working without you. The account model behind all of this is on Agentic AI Users.

Step 1 — Give the Agent an account

Invite the Agent under Personnel, the same way you'd invite a person; an ordinary member account is all it needs. For this walkthrough, the Agent's account is named Abel.

Sign in as Abel and mint its personal access token under Edit Profile → API Access. Agent access must be enabled for the organization (Settings → Integrations, the same switch as always). Then configure Abel's MCP connection with that token, exactly as described on Connecting Your Agent. From here on, everything Abel does over MCP is recorded as Abel.

Step 2 — Tell it what its job is

The Agent's side of the arrangement lives in its standing instructions, in whatever host it runs in:

You have an AbleTime account of your own. At the start of each session, check
the tasks assigned to you and work them one at a time. Record your time as you
go, keep the task description current, and when the work is on the record,
mark the task done, comment with what you did, and @mention gps.
If you need a decision, comment with the question, @mention gps, mark the task
blocked, and move on to the next one.

Step 3 — Hand it a task

Create a task the way you always do, and assign it to Abel:

AAP-50 — Write the release notes for 2.3
Assigned: Abel
Category: Build    State: todo
Description: Draft the 2.3 release notes from the merged changelog.
Cover the three headline features; keep the register plain.

That assignment is the whole handoff. On Abel's next session it finds AAP-50, starts the work, and records time as it goes: draft entries under its own name, each stamped with the tool that wrote it.

Step 4 — Watch the work come back

You see the work the same way you'd see anyone's. AAP-50 moves into doing as Abel's recorded time lands against it, its description grows into a record of what was done, and when the work is on the record, the task is marked done and a comment arrives:

Abel commented on AAP-50:
"Release notes drafted for all three headline features and committed to the
docs branch. Register kept plain per the task description. @josey_b ready for
review."

The @mention notifies you directly, so completion reaches you without polling the board. In the ledger, Abel's accepted time sits alongside everyone else's; reports, billing, and audits see the task, the minutes, and Abel's identity behind them.

Step 5 — When it needs you

A worker that can't proceed doesn't guess. Given the standing instructions above, a stuck Abel parks the task and asks:

Abel commented on AAP-51:
"The changelog names a fourth feature that isn't in the release scope doc.
Include it in the notes or hold it for 2.4? @karis_v"

The task is marked blocked with that question as its note, so the board shows exactly where agent work is waiting on a human. Answer in a comment and unblock the task; Abel picks it back up on its next pass through its assignments.

Step 6 — Put it on a lane

Assignment by hand works; a lane works better once the arrangement is routine. A board lane can serve as Abel's queue:

Watch the "Agent Builder" lane on BuildIt Agentic Projects. When a task appears there,
pick it up, complete it, record your time, comment with what you did, and
@mention karis or tony. Blocked tasks: comment the question, mark blocked, keep going.

You feed the lane; Abel drains it. Moving a card into the lane is the whole handoff, and the lane's card count is the state of the queue.

Reviewing an Agent's record

A person's drafts stay private until accepted; Abel's record carries no such expectation, because its work exists to be reviewed by the people it works with. Review it the way you'd review the board: the tasks say what was done, the comments say what Abel found, and the blocked cards say where it needs you. The tools Abel works with are the same ones on MCP Tools.