Today I finally understood why people keep comparing Pi’s /tree
to Claude Code’s /resume
.
Pi’s official post says sessions are stored as trees. You can jump to any earlier point and continue from there. All branches live in one filterable file.
That sounds a lot like “go back to an old conversation,” so /resume
is the first Claude command that comes to mind. The jobs are related. They are not the same.
What Each Command Does #
| Command | What it does |
|---|---|
Pi /tree |
Browse the current session’s history tree. Pick an earlier message. Continue from there. |
Claude /resume |
Open the session picker and switch to another saved conversation. |
Pi /resume |
Pi’s own session picker. Closest match to Claude /resume . |
Claude /branch |
Copy history into a new session ID. Closest match to Pi /tree ’s “try another path” feel. |
The key insight: /tree
navigates inside one session. /resume
switches between sessions.
Why That Matters #
Pi stores each session as a tree, not a single straight line. Each entry has an id
and parentId
. So one session can hold two explorations:
Initial architecture discussion
├─ Redis route → implementation notes
└─ Postgres route → benchmark plan
You can jump back to “Should we use Redis or Postgres for agent memory?”, pick that node, and continue with a different premise. The Redis path stays. You did not throw it away. You did not open a second chat file either.
Claude Code favors separate session artifacts for branches:
Original session transcript
└─ /branch → separate copied transcript/session
You then use /resume
to move between those files. Useful when an experiment should become its own unit of work. Heavier when you just want path B for five minutes.
Quick Mental Model #
| Need | Use |
|---|---|
| Try path B from an earlier turn, keep path A | Pi /tree |
| Reopen yesterday’s conversation | Pi /resume or Claude /resume |
| Make an experiment its own session | Claude /branch , or Pi /fork / /clone |
One nuance: Pi also has /fork
and /clone
when you do want separate session files. /tree
is the lightweight, same-file alternative.
What I Learned #
- Pi
/tree
is git-like branchinginside one session file - Claude
/resume
is session-level navigation, not an in-session timeline browser - The true Claude pair for Pi
/tree
is closer to/branch
+/resume
- For exploratory design work,
/tree
is lower friction: jump back, try path B, jump back to path A - For independent experiments you want to keep separate, Claude’s copied-session model is clearer
References #
Comparing coding-agent session models? Tell me which workflow clicked for you. Reach out on LinkedIn.