SWE-bench Verified, live, four arms: 13% cheaper than no compaction, and the highest reward. See the numbers

Home and status

Question: is it working, how am I doing, and what should I do next?

/context-guru:status is the home screen. It first checks that this session is routed through the proxy and that the proxy answers, then prints:

  1. one line with the dollars saved over the last 7 days and the share of what the same work would have cost;
  2. today against yesterday, saved and spent, with a trend for the week;
  3. the top three things you could do, each with the dollars at stake.

/context-guru:status /context-guru:status

Saved $41.20 in 7 days (4.6% of what the same work would have cost).

            saved     spent
Today        $6.35   $141.20
Yesterday    $5.10   $118.40
7 days     $41.20   $856.85   saved ▇▁▁█▇▇█  spent ▇▂▁██▇█

≈ Estimated   (spend you could still avoid; rows overlap, do not add)
  1.  $107.79  You paused longer than the cache lasts, 214 times: run /compact before a long break
  2.   $37.70  Add-ons you never use: remove them, /context-guru:waste lists them
  3.   $23.31  Settings or tools changed mid-session: pick MCP servers and edit CLAUDE.md before you start

/context-guru:savings  /context-guru:cache  /context-guru:waste  /context-guru:session  /context-guru:doctor
/context-guru:watch  /context-guru:timeline  /context-guru:chart  /context-guru:heatmap  /context-guru:fill

Reading it

When it says something is wrong

The routing check tells apart the common cases: routing written but this session started before it (normal right after install; it applies to the next session), the proxy not answering, a base URL that is not this proxy's, and a proxy started with older settings than you have now. For anything else run /context-guru:doctor.

/context-guru:status --stats prints the proxy's raw /stats JSON and nothing else.