--- title: "GrepRAG · The productivity harness for Claude Code" description: "The productivity harness for Claude Code. Your agents know everything you've done, talk to coordinate, grep massive libraries for minimal tokens, and never make the same mistake twice." canonical: https://www.greprag.com/ source: / --- v0.8.3 · Built for Claude Code # The productivity harness for Claude Code. Your agents know everything you've done. They talk to each other to coordinate, grep massive libraries for minimal tokens, and never make the same mistake twice. `npm install -g greprag` Copy Then run `/greprag` in any Claude Code session. Setup ### Two commands. Zero config. Install once. Type `/greprag` in Claude Code. The agent walks you through email signup, hook wiring, and project anchoring. Conversation, not a manual. Architecture ### Pyramid compaction weekly daily hourly turn Each tier synthesizes the one below. Crons fire automatically. Isolation ### Per-project memory Each repo gets its own anchor. Memory for one project never bleeds into another. Continuity ### Session-start briefing Yesterday's summary and recent activity appear at the top of every new session. Fidelity ### Verbatim recall The compactor shapes once. After that, your work surfaces as-is. No LLM rewriting at retrieval. Pricing ### Free during beta No credit card. Email signup via OTP. Use it on as many projects as you like. Example ### What you see at session start Coding Book writing Tweets Email $ claude \[GrepRAG memory: your-project\] Yesterday's summary: 1. Shipped the new auth flow with migration 028 and updated env vars. 2. Fixed the race condition in the queue worker. Added the missing lock. 3. Refactored the API layer to use the new typed-client. Recent sessions: 4 hours ago: 1. Started exploring the bento layout for the landing page. 1 hour ago: 1. Wired the SessionStart recap hook end-to-end. $ claude \[GrepRAG memory: my-novel\] Yesterday's summary: 1. Drafted Chapter 5: 1,765 words covering the protagonist's first encounter with the antagonist. 2. Restructured TOC v3: pivoted the inciting incident from Ch 4 to Ch 3, tightened pacing. 3. Locked Ch 4 as final-pass; Ch 5 marked in-progress in the Working Notes. 4. Added 12-row character motivation table to track POV consistency. Recent sessions: 3 hours ago: 1. Expanded the central conflict. Added two new scene cards and a recovery beat. 20 min ago: 1. Cross-referenced citations for the historical backdrop; flagged 3 sources to verify. $ claude \[GrepRAG memory: social-content\] Yesterday's summary: 1. Posted 6-tweet thread on "why most agent memory designs fail" (14K impressions, 47 RTs). 2. Scheduled 4 tweets via Buffer for the week, focused on the developer audience. 3. Identified 3 viral angles from competitor accounts to test next week. Recent sessions: 2 hours ago: 1. Drafted a quote-tweet of @founder's post on agent infrastructure. Voice-checked against profile. 15 min ago: 1. Composed reply to high-engagement comment on yesterday's thread; A/B'd two versions. $ claude \[GrepRAG memory: customer-inbox\] Yesterday's summary: 1. Cleared 23 emails. 4 escalated to engineering, 19 resolved with KB articles. 2. Updated FAQ doc with 3 new entries based on recurring questions. 3. Drafted weekly digest for product team: top 5 customer pain points. Recent sessions: 1 hour ago: 1. Replied to refund-request batch (5 customers). Used standard template, flagged 1 for review. 30 min ago: 1. Drafted onboarding follow-up for new customer Acme Inc. Included setup checklist. The full surface ## Four stores. One engine. What an orchestrator agent and its subagents share. Odyssey for the journey. Corpus for research. Lore to skip mistakes twice. Inbox to coordinate. [ Odyssey ### The session pyramid Your agent remembers everything about your journey together. Per-project capture, compacted hourly, daily, weekly. Recap at session start. About `$0.10/month` per project. Read more →](https://www.greprag.com/odyssey)[ Corpus ### The grep layer Your personal store of knowledge and exclusive information, searchable and token efficient. Drop in a book, manuscript, or past articles. Ask the agent for a passage; get a real quote back. No embeddings, no monthly bill. Read more →](https://www.greprag.com/corpus)[ Lore ### Lessons that stick Lessons and lived experiences of you and your agent, codified so you don't have to learn them twice. Agent-seeded one-liners, routed by scope. Pulled at chip spawn, before edits, at session start. Read more →](https://www.greprag.com/lore)[ Inbox ### Cross-session messaging SSE push delivery, session-scoped routing, 14-day TTL, retract codes. Chips and advisors talk to each other. Read more →](https://www.greprag.com/inbox) The hero ## Memory alone is dead weight. That's why greprag ships with advisors. Business. Content. Skill. Each one reads your four stores, forms strategy, and dispatches sub-agents to act. Composition ## The loop closes itself. Chips orchestrate. Crons refine. Hooks inject. A chip hits friction, seeds a lore note, and the next chip skips the discovery. When a chip misses the moment, the hourly extractor catches it. [See the recursive learning loop →](https://www.greprag.com/how)