Outputs & Resume

Arbor records everything a run produces so you can inspect it, reproduce it, and pick up where you left off.

Session artifacts

Each run gets its own session directory, by default under the target project:

<project>/.arbor/sessions/<run_name>/

<run_name> defaults to a timestamp; set it explicitly with --run-name, or relocate the whole directory with --workspace-dir.

Inside you'll find the run's checkpoint (Idea Tree + message history), logs, and the final report. The exact instruction you launched with is recorded in the session log, and the fully-resolved settings for the run are saved to <run_name>/.coordinator/config_snapshot.yaml (every config layer merged, secrets redacted).

Reuse a setup

To repeat a study with the same settings in a fresh run, copy that config_snapshot.yaml to your project root as arbor.yaml — Arbor auto-loads it next time. See Preparing a Benchmark → Saving and reusing a setup.

REPORT.md

When a run finishes, Arbor writes a REPORT.md — the human-readable write-up of the study: what was tried, what worked, the evidence behind each conclusion, and the final result. Use arbor report to work with it, and the /report slash command during a run to print artifact paths.

By default Arbor then opens a read-only Q&A prompt so you can interrogate the finished run (disable with --no-followup).

Experiment branches

Every experiment ran on its own git branch in an isolated worktree. Merged improvements are on trunk; explored-but-unmerged ideas remain as branch refs you can inspect. During a run, /branches lists the explored branch refs and /tree prints the Idea Tree snapshot.

Saving & resuming the planning conversation

The intake chat — what you do at the arbor prompt before a run launches — is itself saved every turn, so you can step away and pick it back up. Each conversation lives in its own directory under the launch folder:

<cwd>/.arbor/conversations/<conv_id>/

with the message history (messages.jsonl) and a small meta.json (title, timestamps, turn count). Saving is best-effort and never interrupts the chat.

Two ways back in:

• arbor --continue (-C) — reload the most recent unfinished conversation and keep chatting, like claude -c. (-c is already --config, so the short flag is -C.)
• /resume inside intake — lists saved conversations (tagged [chat]) alongside launched runs; pick a conversation to keep chatting, or a run to replay the engine.

On resume the prior exchange is replayed to the terminal so you can see what was already said. Once a conversation launches a run it is recorded as launched and dropped from --continue — the run's own checkpoint (below) takes over.

Resuming an interrupted run

If a run is interrupted — you stopped it, the machine rebooted, a budget tripped — resume from its checkpoint instead of starting over:

arbor --resume --run-name my-study

--resume reloads the Idea Tree and message history from the existing session and continues in the same workspace. Combine it with --run-name (or --workspace-dir) to point at the session you want to continue.

Tip

Because the Idea Tree is the durable shared state, resuming restores not just where the run stopped but what it had learned — merged improvements, pruned branches, and backpropagated insight are all intact.

Monitoring while running

• Terminal dashboard — live cycle status, Idea Tree, and cost.
• Read-only web monitor — auto-starts near port 8765 (--webui-port to change, --no-webui to disable).
• /cost — print token usage at any time.