Log inSubmit server
Crbro Memory logo

Crbro Memory

Persistent, local MCP memory with searchable facts, session context, corrections, credential handling, and optional team sharing.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Crbro Memory

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "crbro-memory": {
      "command": "npx",
      "args": [
        "-y",
        "crbro-memory"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Install not yet sandbox-tested— the server itself is confirmed live (we called its tools/list endpoint). The stdio install command is a separate, ongoing check.

Overview

The crbro-memory MCP server gives AI assistants persistent local memory across conversations. It stores facts, decisions, patterns, preferences, errors, debts, context, and session summaries in a file-based brain organized into neurons and synapses. Use its read and write tools to recall earlier project knowledge, correct or remove outdated information, manage credentials through the operating system keychain, and optionally share selected project neurons through Git-backed team spaces. Reach for it when an agent needs inspectable, versionable memory without sending the brain to a hosted service.

Use cases

  • Recall project decisions from earlier conversations
  • Store and update durable agent facts locally
  • Track open work items across sessions
  • Audit and isolate credentials from persistent memory
  • Share selected project knowledge with teammates

Key features

  • Persistent file-based memory
  • Fact-level keyword and optional semantic search
  • Inspectable neurons, entries, maps, and sessions
  • Superseding, revising, and destructive forgetting workflows
  • Operating-system keychain credential storage
  • Explicit Git-backed team-space sharing

Tools15

What Crbro Memory lets your agent do. Expand a tool for its parameters.

Verified live~7.8k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

12/14 checks healthy over the last 92d

What it can do

From the 15 tools this server returned when we ran it; 15 declare their own safety hints.

Can delete or modify data · 7Handles credentials · 7Makes network requests · 5Read-only tools · 8
Network at startup3 connections
Files written at startupNone
Install scriptsNone
npm provenanceNot published
Package age161 days
Maintainers1

Observed 2026-10-15 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 7d ago via OSV.dev · crbro-memory (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars8Stargazers on the source repository.
npm downloads1kPackage downloads in the last 30 days.
Last commit23d agoMost recent push to the default branch.
Tools exposed15Callable tools this server registers over MCP.
Directory activity1 viewsConfig copies, upvotes, and views on AllMCPs.

How Crbro Memory works

What crbro-memory does

The crbro-memory MCP server gives an AI assistant a persistent memory store that remains available across sessions. It is designed for local use: knowledge is kept in a file-based brain, with neurons for topics, synapses for relationships, and session logs for conversation summaries. The brain can contain facts, decisions, patterns, preferences, errors, debts, system maps, and working context.

This fits the Knowledge & Memory category for agents that need continuity across projects or conversations. Memory is inspectable and editable rather than being an opaque history. The tools distinguish between information that has changed, information that should be removed entirely, and information that is still valid but should be verified.

How it works

A typical conversation begins with crbro_boot, which loads earlier context, prepares the search index, and places project-specific neurons first when a project is supplied. Before answering questions about prior work, preferences, decisions, or projects, the agent can use crbro_recall. Results identify the matching entry, its kind, date, confidence, and possible freshness concerns. crbro_inspect then reads a neuron or selected entries without changing the brain.

New knowledge is saved with crbro_learn. Repeated text is not blindly duplicated, while changed confidence, shelf life, and confirmation state can be updated. Replacements can supersede earlier facts; knowledge that is no longer true can be retired with crbro_revise; information that must leave disk uses crbro_forget. crbro_consolidate closes a session, records its summary, persists pending knowledge, recalculates heat, and links neurons touched during that session.

The crbro-memory MCP server can also maintain a working context with open and recently closed items. System maps provide a single prose description of where a system lives and how its parts connect. Optional semantic recall uses a local embedding model alongside keyword search.

Setup and configuration

The package is available through npm, and the documented initialization command is npx crbro-memory init. Initialization installs the local multilingual-e5-small embedding model by default; init --no-semantic skips that installation. Semantic search can also be disabled with CRBRO_SEMANTIC=0. The semantic model uses about 500 MB of disk space once per machine and about 0.5 GB of RAM while a server is running.

CRBRO_SHELF_DAYS configures default shelf-life durations for facts. CRBRO_BACKUP_DIR places daily compressed backups in a specified synced folder. An optional daemon can let multiple clients use one process and shared index through npx crbro-memory daemon on.

Tools and capabilities

The crbro-memory MCP server exposes tools for:

  • Booting, recalling, inspecting, and consolidating memory.
  • Learning, revising, forgetting, and reconnecting stored knowledge.
  • Managing active context, system maps, heat scores, synapses, and maintenance.
  • Auditing stored content for credentials without revealing their values.
  • Storing credentials in macOS Keychain, Linux Secret Service, or Windows DPAPI through crbro_secret.
  • Sharing selected neurons through private Git-backed team spaces.

Maintenance can rebuild the index, check integrity, prune weak connections, archive cold neurons, and report debts without revisit triggers. Sharing is explicit per neuron; preferences remain local, and credentials are refused from team sharing.

Limitations and notes

The memory is local and does not provide access control: project scope organizes recall but is not a security boundary. Semantic recall requires additional disk and runtime memory, while disabling it leaves keyword search available. Expired facts are marked as possibly stale and require source checking; relevance and confidence labels are not guarantees of truth.

Session knowledge is logged only when crbro_consolidate is called. Leaving a session without consolidation can lose its session knowledge. Deleting credentials from memory does not rotate the credential itself. Shared content already sent to a remote remains there after unsharing. Team synchronization can be offline; pending notes remain unsent until a later sync.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Crbro Memory

crbro-memory is an MCP server that connects an AI assistant to a local file-based memory brain and optional Git-backed team spaces. Its main tools boot, recall, inspect, learn, revise, forget, consolidate, and manage secrets, supporting persistent knowledge, corrections, context, sessions, credentials, and selected project sharing across conversations.

Run `npx crbro-memory init` to initialize it. Initialization installs the local semantic model by default; use `npx crbro-memory init --no-semantic` to skip that model.

Yes, Cursor is named as a compatible MCP client in the project documentation. The material also names Claude Desktop and Claude Code. It does not provide additional Cursor-specific setup instructions.

Use `crbro_secret` to store credentials in the operating system keychain instead of the brain. The memory records only a credential name or marker, and auditing reports credential locations and types without exposing values. Credentials must still be rotated after removal.

Create or join a team space with `crbro_space`, then explicitly share selected neurons with `crbro_share`. Shared notes synchronize through a private Git repository. Preferences remain local, and sharing refuses content containing credentials.

Related MCP servers

All in Knowledge & Memory All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewCrbro Memory AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/crbro-memory?style=directory)](https://allmcps.com/mcp/crbro-memory)
HTML Embed
<a href="https://allmcps.com/mcp/crbro-memory"><img src="https://allmcps.com/api/badge/crbro-memory?style=directory" alt="Crbro Memory on AllMCPs" /></a>