# Good agents have todo lists, great agents have todo graphs

Canonical: https://tiptreesystems.com/blog/todo-graphs

2026-09-07 · Team Tiptree

Good agents have todo lists, but great agents have todo graphs. This month we improved our todo engine (which is live, you can try it in Althea).

You don’t need to know any of this, it’s just how Althea keeps track of the work it’s doing. But we think it’s cool so we thought we’d share.

Now, our todo items declare their dependencies, so the plan is a directed acyclic graph (DAG) where tasks are block until dependencies resolve. The plan also supports conditional branching: If A -> B, but if A’ -> B’. This is helpful when the agent needs to make plans that involve getting more information (which is often the case!). When a branch is selected, the todo engine cancels the losing branches in a cascade, so the agent can’t end up crossing off cancelled tasks.

Importantly, a person (or another agent) can edit a running agent’s plan. Edits are applied as jobs serialized against the agent’s wake cycle, so an edit never races the agent mid-thought.

The picture below is a real tree from one of our own Althea sessions on August 19. A coordinator agent asked a question; Althea decided to bring its user in rather than answer alone; an outbound reply then waited on an explicit consent decision, and the user asked for edits, so two branches died.

The plan’s rules are enforced by our Tiptree's runtime - this makes the agent far more reliable and able to carry out complex tasks.

This is Engine Room / 001, notes on the machinery behind Althea and Lacuna.

*First posted on [LinkedIn](https://www.linkedin.com/feed/update/urn:li:activity:7502758901566754816), September 7, 2026.*
