Skip to main content
Glama

Verbative Memory

On-device memory system for AI coding agents — a dual-layer engine (distilled facts + a lossless event ledger), automatically captured and injected, exposed over MCP. Shared across agents, persistent across sessions; stored as plain, human-readable files in your repo — no memory cloud, no vendor database.

This repository is the public home for docs, install help, and issues. The engine is closed-source and ships as a sealed, compiled package — there is no source code here.

How it works — capture everything, recall selectively

  • Complete capture, in two layers. Underneath sits a lossless ledger: a mechanical, no-model record of every command, edit, and test result, so "what did we run" and "when did this test start failing" are exact, replayable facts. On top, an on-device model distills that stream into atomic, typed memories — decisions, conventions, and dead ends.

  • Selective recall — the two layers fused. A single recall surfaces only the handful of distilled facts that matter (ranked by semantic search and a reranker) and interleaves them, by time, with the exact ledger record around them: what you decided next to what you actually did. One answer, not a dump of the whole history.

  • Self-organizing. New facts reconcile against old ones, importance decays on per-type forgetting curves, frequently-recalled facts strengthen, and recurring specifics distill into higher-level insights.

  • Yours, on-device. Memories live in human-readable files in your project. Recall makes zero cloud calls.

Related MCP server: exocortex

Requirements

  • macOS (Apple silicon) or Windows 10/11 (x64). (Linux is on the roadmap.)

  • A Verbative account — Memory is part of the Code Advanced plan and starts with a 14-day free trial. See the plans and the Terms.

Install

Claude Code CLI — automatic capture and recall:

claude mcp add verbative-memory -- npx -y verbative-memory

pip:

pip install verbative-memory
verbative-memory login      # sign in with your Verbative account

Any MCP client (mcp.json / stdio):

{ "mcpServers": { "verbative-memory": { "command": "npx", "args": ["-y", "verbative-memory"] } } }

Automatic capture and injection are built for the Claude Code CLI and Codex; any other Model Context Protocol client can call the tools manually. The first run downloads the on-device engine and models once; after that, recall runs fully offline.

More detail in docs/INSTALL.md and at verbative.de/docs/memory.

Tools

remember · recall · search · ask · list · get · pin · forget · revise · edit_core · feedback · related · history · timeline · events — and more. See the docs.

Privacy

Capture, embeddings, reranking, and the knowledge graph all run locally. Your code and memories are not uploaded; the memory files stay in your project and leave your machine only if you choose to share them.

Issues

Bug reports and feature requests are welcome in the issue tracker — the place for anything about the docs or the tool. For your account, licensing, or a trial, email info@verbative.de.


Verbative is a product of Cramer Digital. Not affiliated with, or endorsed by, Anthropic. Verbative works alongside Anthropic's Claude and the Claude Code CLI; those names belong to Anthropic. See LICENSE.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      Personal unified memory system for AI coding agents, providing persistent memory with hybrid RAG retrieval via MCP integration, allowing agents to store, search, and manage memories locally.
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      A persistent, local memory layer for AI coding agents that remembers decisions, bugs, and rules across sessions with three core MCP verbs (recall, remember, search).
      Apache 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Local-first hierarchical memory for AI coding agents, providing automatic session capture, belief consolidation, and relevant memory injection via an MCP server.
      70 npm
      2
      MIT