Engrim

What hackers pick instead

• MemPalacePeople recommend it as a locally managed memory layer for AI agents that retrieves past conversations well. Critics say independent end-to-end tests scored far below its retrieval benchmarks.9recommend
• GrovPeople recommend it for cutting exploration tokens and recalling past sessions for Claude Code. One user says it is rough around the edges, and it needs an API key.8recommend
• claude-memRecommended as an open-source Claude Code memory plugin that recalls past prompts and replies without many tokens. One caveat: it only records from install forward.8recommend
• LettaFree open-source agent framework with selective memory retrieval and long-context handling. Critics point to heavy dependencies and embedding API costs.8recommend
• Hippo AI agent memoryRecommended for modeling memory consolidation and decay for AI agents rather than a plain vector store. Critics question whether decay is needed.8recommend

Every quote

The 29 most detailed of 31 comments. Each links to the original.

1. Instead of vector DB + semantic search, it uses models from cognitive science (ACT-R activation decay, Hebbian learning, forgetting curves)

   Recommends · Mar 2026 · in “Ask HN: What Are You Working On? (March 2026)”

2. tracking which agent (or human) made which decision ended up being one of the most useful parts of the whole system for me

   Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

3. separating settled decision records from WIP state tells the new agent what was finished vs what died mid-flight

   Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

4. modeling things like salience, decay, and consolidation is much closer to how useful memory actually works

   Recommends · Feb 2026 · in “Show HN: Claude-engram – Brain-inspired persistent memory, runs inside Claude.ai”

5. The result is agents that remember context across sessions, learn from patterns, and forget gracefully.

   Recommends · Mar 2026

6. Routine operations evict from a circular buffer. Errors, breakthroughs, and novel work persist.

   Recommends · Mar 2026 · in “Engram – salience-gated memory for Claude Code (captures what matters)”

7. agents that remember context across sessions, learn from patterns, and forget gracefully

   Recommends · Mar 2026 · in “OpenPawz Engram biologically-inspired memory architecture for AI agents”

8. It's an improvement, but obviously you're still limited with effective context size

   Mixed · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

9. an MCP server that gives all your agents a shared, persistent knowledge base

   Recommends · Apr 2026 · in “I built a multi-agent memory consistency layer after the Claude Code leak”

10. without requiring a vector database, an OpenAI key, or any external service

    Recommends · Mar 2026 · in “Show HN: Engram – open-source persistent memory for AI agents (Bun and SQLite)”

11. The consolidation and contradiction detection features are smart additions

    Recommends · Mar 2026 · in “Show HN: Engram – Persistent memory for AI coding agents (open source)”

12. Local SQLite is practically zero-latency, keeps your IP entirely private

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

13. I've been using for months to allow LLM agents to have task management.

    Recommends · Mar 2026 · in “Engram – A distributed memory system for AI agents, with extensible architecture”

14. Local-first + SQLite is the right call for offline-first agent memory

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

15. local embeddings without needing an API key is the right call

    Mixed · Mar 2026 · in “Show HN: Engram – open-source persistent memory for AI agents (Bun and SQLite)”

16. Single binary, single .brain file, no external dependencies.

    Recommends · Mar 2026 · in “Engram – AI memory engine: knowledge graph, search, reasoning in one binary”

17. I especially like the provenance idea.

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

18. Decision summaries look really cool

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

19. I like this local-first concept.

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

20. Engram uses three strategies in parallel: vector embeddings (nomic-embed-text via Ollama, local-first), BM25 keyword, and temporal recency

    Recommends · Mar 2026 · in “Anchor Engine – deterministic semantic memory for LLMs, <1GB RAM runs on a phone”

21. Because the Codex CLI supports the Model Context Protocol (MCP) and lifecycle hooks, it actually worksperfectly out of the box.

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

22. the real value now with engrim is if you are using multiple services, your project is relevant when switching services

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

23. Nice tool you shared, feel free to mess around with engrim and provide feedback. You may end up really liking it.

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

24. Engram already has namespace isolation — API keys scope memory per-agent, spaces partition further within a user

    Recommends · Mar 2026 · in “Show HN: Engram – open-source persistent memory for AI agents (Bun and SQLite)”

25. replaces attention dilution with a 4,000-character curated episodic working memory pack

    Recommends · Sep 2026 · in “Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs”

26. lets you reduce your token burn rate by a factor of huge magnitudes

    Recommends · Apr 2026 · in “Do you ever ask "Please Claude I need this my account is kinda tokenless "”

27. the problem is a model problem not a systems problem

    Recommends · Apr 2026 · in “The M×N problem of tool calling and open-source models”

28. Recall scoring got noticeably better after this.

    Recommends · Mar 2026 · in “Show HN: Engram – open-source persistent memory for AI agents (Bun and SQLite)”

29. More useful for RAG/agent workflows.

    Recommends · Mar 2026 · in “Show HN: Engram – open-source persistent memory for AI agents (Bun and SQLite)”