Sessions

Dynamo auto-saves conversations to local storage so you can pick up where you left off whenever the session file is still on disk.

How Sessions Work

Every exchange is saved automatically to the sessions store in your platform config dir. Sessions capture:

• ●Messages (user and assistant)
• ●Model and provider used
• ●Token usage and running cost
• ●Timestamps
• ●Tool permissions granted (per model, so "Allow every" survives a resume)
• ●Your task list
• ●Reasoning effort and sandbox mode

Session Names

Sessions are named automatically after the first reply, from what you asked. A name you set is never replaced. Rename anytime:

/rename Camera System Refactor

Resuming

# Interactive picker
dynamo -r

# Or from within Dynamo
/resume

/resume defaults to sessions from the current project (the directory you launched Dynamo in), so you only see relevant conversations. Pass a session id directly with dynamo -r <id> to jump straight to one.

Context Management

Dynamo tracks context window usage with color-coded indicators:

Auto-Compaction

When the context window fills up, Dynamo automatically compacts — it summarizes older messages and keeps the most recent ones word for word. The full conversation from before the compaction is saved under .dynamo/cache/transcripts/, and the model can read it back if it needs a detail. Your message is still answered; nothing is dropped. Press Ctrl+O to expand or collapse the summary. Toggle auto-compact in /config.

On a Claude or Grok subscription, the subscription manages its own context, so auto-compaction doesn't run. /compact still works.

Dynamo won't switch to a model whose window would be more than 90% full.

Manual Compaction

/compact
/compact keep the shader debugging details

Add a focus to steer what the summary keeps.

Starting Fresh

/clear

This starts a new session. The previous one is kept for /resume (empty sessions aren't saved). To branch instead, use /fork — it copies this session up to a reply into a new one.