How do i access if it's working or not ?

[Shreyaaaaaak]

How do i access if it's working or not ? I mean i dont know how to spot the difference cause i'm a beginner so if anyone can help me out that would be great

[maxiao1204]

clone one repo in two folders and then use same promt in both folder and one use code review graph while the other dont use it, then check the useage..

[tirth8205]

Three ways to spot the difference, all local (v2.3.5):

1. code-review-graph detect-changes --brief — prints a Token Savings panel (baseline vs graph context vs saved); add --verify to cross-check the estimate against a real tokenizer.
2. code-review-graph status — node/edge/file counts and freshness; if this looks wrong, the graph isn't what your agent is querying.
3. In your AI tool, check the MCP server is actually connected (Claude Code: /mcp) and that the session transcript shows graph tool calls — if the agent never calls them, you get no benefit regardless of config.

The new docs/FAQ.md has a "verify it's working" section, and a one-shot doctor command is planned. Note the A/B experiment suggested above was run by another user in #467 with a negative result — I've replied there with the verified causes; worth reading before relying on instructions alone.