AI session continuity

A token limit shouldn't wipe out a day of work.

A reliable resume starts with the execution environment, then loads the decisions and the useful messages. It doesn't rely on a vague summary pasted into a new chat.

Resume checklist

What the new agent must know before it writes.

Where does the work happen?

The repository, folder, branch, worktree, and any uncommitted changes come first.

What outcome is expected?

The goal, the scope, and the acceptance criteria prevent a technically correct but off-target solution.

What has already been decided?

Architecture choices, trade-offs, and explicit refusals keep the new agent from reopening the debate.

How do you verify?

Tests run, remaining errors, validation commands, and known risks close the loop.

Resume and memory

Is a MEMORY.md file enough?

It helps a lot, but it doesn't replace the project path, the Git state, the source messages, and the execution traces. Tramya links these elements rather than picking a single memory.

Can you resume in the same agent?

Yes. When the source offers a native resume, Tramya prefers it. Otherwise, it opens a new session with a handoff brief.

Why not send the entire history?

A massive history eats up the context window and buries the important decisions. Targeted search and a compact handoff generally give a better resume.