<aside>

All our agents support the shared Context Checkpoints system.

Context Checkpoints help you:

• Preserve the current state of a conversation, including relevant context, decisions, constraints, open questions, and next actions.
• Record what has changed since the previous checkpoint, creating a clear reference for future work.
• Continue in a new thread without starting again from zero, using the latest checkpoint as the current source of truth. </aside>

<aside>

🟡 What is a Context Checkpoint?

</aside>

A Context Checkpoint is a structured summary of the relevant state of a conversation at a specific moment.

It brings together the information needed to understand where the work currently stands, such as:

• current context
• decisions made
• active constraints
• open questions
• risks
• important changes
• next actions

If no previous checkpoint exists, the checkpoint reviews the conversation from the beginning and captures the relevant current state.

If a previous checkpoint exists, the new checkpoint uses it as the starting point. It preserves the information that is still valid, adds or updates the relevant information from the conversation since then, and removes or corrects information that is no longer useful or accurate.

In practice:

*previous checkpoint
+
relevant information added or changed since then
-
information that is no longer valid
=
one new, complete, self-contained checkpoint*

A checkpoint does not reproduce the full conversation or preserve every historical detail. It keeps only the information needed to understand the current state and continue the work.

<aside>

🟢 What is it useful for?

</aside>

A Context Checkpoint can help you: