Praetorian FAQs

Question 1

What is `praetorian_compact` and when should I use it?

Accepted Answer

`praetorian_compact` is a command to incrementally compact context into structured TOON snapshots. You should use it frequently: after web research, task completions, reading multiple files, making decisions, or during long conversations to proactively save tokens and refresh context.

Question 2

Does Praetorian handle duplicate or similar context?

Accepted Answer

Yes, Praetorian intelligently handles similar context. It uses Jaccard similarity to detect when new compaction is over 70% similar to an existing one, automatically merging the new content without duplication and updating the existing snapshot.

Question 3

How can I retrieve or search for previously compacted context?

Accepted Answer

You can use `praetorian_restore` to search for and inject relevant context back into your current Claude Code session. It supports keyword queries via an inverted index or can list your most recent compactions.

Question 4

What is Praetorian and what problem does it solve for Claude Code users?

Accepted Answer

Praetorian is an MCP (Model Context Protocol) server for Claude Code that aggressively compacts and recycles context. It solves the problem of excessive token usage by structuring web research, task outputs, and conversations into efficient snapshots, leading to 90%+ token savings.

Question 5

How does Praetorian achieve significant token savings?

Accepted Answer

Praetorian utilizes the highly efficient TOON (Token-Oriented Object Notation) format, which is 30-60% smaller than YAML/JSON. It also employs smart merging, deduplication based on Jaccard similarity, and an inverted index for optimized context management.

Praetorian

Praetorian

Key Features

Use Cases

Key Features

Use Cases