context-guru

Provider-agnostic context engineering for LLM agents.

▶ Watch the demo

---

context-guru cuts the token cost of your agent's traffic in two ways: carry less context (drop redundant tool output, collapse superseded runs, summarize before you hit the limit), and pay less for what you still carry (keep your prompt cache warm, split the volatile tail off the system prompt so the rest stays cacheable). Paying less is the default — it's on out of the box, before you opt into anything that trims content.

Full docs: rossoctl.github.io/context-guru.

Install

Choose your setup. Each button opens only the instructions for that path.

  </td>
  <td align="center" rowspan="2">
    
    Claude Code or Codex
  </td>
</tr>
<tr>
  <th>Codex (experimental)</th>
  <td align="center">
    
  </td>
</tr>

For local Codex use, register only the marketplace paths Codex needs; cloning the whole monorepo can exceed Codex's marketplace timeout:

codex plugin marketplace add rossoctl/context-guru --sparse .agents --sparse codex-marketplace
codex plugin add context-guru@context-guru

Claude Code plugin

The plugin answers three questions in the terminal: how much am I saving, is my cache about to cost me, what should I do?

/plugin marketplace add rossoctl/context-guru
/plugin install context-guru@context-guru

Each command opens with one line that answers its question, then a small picture; add details for the tables.

Command	The question	Example first line
/context-guru:status	Is it working, how am I doing, what next?	Saved $41.20 in 7 days (4.6% of what the same work would have cost).
/context-guru:savings	How much did it save?	Saved $41.20 in 7 days, net of Context Guru's own spend (4.6% of the $898.05 the same work would have cost).
/context-guru:cache	Is my cache about to cost me?	Cache is warm for 2:41 more. If it lapses, the next message costs $1.91 extra.
/context-guru:waste	What do I carry and never use?	6 things you carry were never used: $37.70 in 7 days, about $161.57 a month.
/context-guru:session	What happened in this session?	In the last 4 hours of this session (9h20m long), cache misses cost you about $4.20 of the $13.50 billed (2 misses in 31 messages).
/context-guru:doctor	Is it set up right?	All 7 checks pass.

Five more views draw in the terminal (watch, timeline, chart, heatmap, fill).

Full guide, every status line state and troubleshooting: rossoctl.github.io/context-guru/plugin.

Presets

It's an effort ladder — each tier is everything in the one before it, plus more: