Memorymesh 48b5187b

MCPAI代理

提供方 gokayokutucu

AI代理 未标注 会话登录 社区 / 第三方 Claude DesktopClaude

MemoryMesh是一个本地优先的AI工作流内存栈,提供CLI工具管理Docker运行时环境,支持文档导入和AI代理内存配置。

一句话选型结论

按资料字段整理

Memorymesh 48b5187b 属于AI代理类 MCP 服务器,传输方式未标注,会话登录,可直接接入 Claude Desktop、Claude,没有现成安装命令,需按官方文档接入。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类AI代理 传输方式未标注 鉴权会话登录 部署方式未标注
GitHub 星标 3 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

AI代理
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

AI工作流TypeScript本地部署本地优先内存管理文档导入Docker运行时

安装与配置

未标注 会话登录
上游未给出传输方式
  1. 1
    准备运行环境

    按官方文档确认运行环境(常见为 Node.js 或 Python),装好后再执行安装命令。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
给代理接外部能力

让代理在任务中途调用搜索、数据库等真实工具。

多步任务编排

把一串操作拆成步骤,由代理按顺序执行。

自动化重复流程

把固定流程交给代理跑,减少手工操作。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 已适配 Claude Desktop、Claude,换客户端不用重新研究配置。

先想清楚的地方

  • 没有现成安装命令,需要按官方文档自己填启动命令和参数。
  • 使用前要准备好会话登录对应的凭据,并按最小权限配置。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在AI代理场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Memorymesh 48b5187b当前 未标注 未标注 未标注 无 2026-05-17
Filesystem 560c4363 13 未标注 本地部署 无 2026-05-17
Server Everything (everything) 11 stdio · 本地进程 本地部署 有 2026-05-17
Time 7365cb6d 2 SSE · 服务端推送 远程部署 有 2026-05-18
Everything Claude Code 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Openmemory 557e1585 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

AI代理
字段来自项目仓库与公开索引,中文说明由本站标注
提供方gokayokutucu上游标注的维护方 gokayokutucu
分类AI代理本站按上游分类展示
连接方式未标注上游未给出传输方式
鉴权方式会话登录远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · 未标注本地安装时用到的包标识与运行时
部署类型未标注上游给出的部署形态
最近核验2026-05-17上游最后核验时间

详细介绍

9.6 KB
来自项目仓库与上游目录的原文,未做删改

MemoryMesh

What is MemoryMesh

MemoryMesh is a local-first memory stack for AI workflows. It ships a CLI (memorymesh) that installs and manages a Docker-based runtime (MemoryMesh API, MongoDB, Neo4j, Qdrant, Ollama), plus CLI flows for setup, diagnostics, lifecycle, GPT archive import, and document import.

Install (Recommended)

macOS / Linux:```bash
curl -fsSL https://github.com/gokayokutucu/memorymesh/releases/latest/download/install.sh | bash


Windows (PowerShell):```powershell
irm https://github.com/gokayokutucu/memorymesh/releases/latest/download/install.ps1 | iex

Release-based installer behavior:

  • Installs/updates the memorymesh CLI globally
  • Launches memorymesh after install
  • Attempts best-effort prerequisite installation where supported
  • Prints explicit platform guidance when prerequisites cannot be auto-installed

Alternative manual npm install:```bash
npm install -g @okutucu/memorymesh
memorymesh


Notes:

- Package name is `@okutucu/memorymesh`
- First run launches interactive setup if `~/.memorymesh/config.json` is missing
- Published tarball includes required internal runtime modules for standalone install

Prerequisites (current phase):

- Node.js 18+
- npm in `PATH`
- Docker Desktop / Docker Engine

Important:

- Bootstrap scripts do not guarantee universal automatic prerequisite installation
- Release assets `install.sh` and `install.ps1` are produced from `scripts/bootstrap/`

## Runtime Modes

Current product state:

- `docker` mode: active and supported
- `native` mode: visible in roadmap, currently disabled / not implemented

## First Run Setup

When no install config is present, running `memorymesh` starts setup wizard flow:

1. Docker checks (`docker --version`, `docker info`)
2. Embedding selection (`flash` or `medium`)
3. Runtime config persistence under `~/.memorymesh`
4. Stack startup with Docker Compose
5. Ollama readiness + model pull
6. Health verification
7. Optional Claude Desktop MCP integration

Installer-managed runtime home:```text
~/.memorymesh/
  config.json
  runtime.env
  stack/
    docker-compose.yml

Embedding Configuration Authority

MemoryMesh uses a single source of truth for embedding settings:

~/.memorymesh/config.json -> generated ~/.memorymesh/runtime.env -> runtime execution

Rules:

  • Change embedding mode/model by re-running setup (memorymesh)
  • Advanced users may edit ~/.memorymesh/config.json directly
  • runtime.env is generated automatically from config.json
  • If drift is detected, MemoryMesh regenerates runtime.env
  • Do not modify embedding via environment variables in normal usage; these values are managed automatically by MemoryMesh

Advanced note:

  • Embedding env vars can exist for internal transport/debugging, but they are not a supported user configuration surface

Core Commands

Command What it does When to use Caveats
memorymesh Runs first-run setup if needed, otherwise opens runtime menu Normal entrypoint Requires working Docker stack for runtime actions
memorymesh doctor Runs diagnostics and prints PASS/WARN/FAIL Health checks and triage Returns non-zero on critical failures
memorymesh doctor --fix Attempts safe automatic repairs, then rechecks Repair common issues quickly Conservative fixes only; may still require manual action
memorymesh start Starts managed Docker stack Bring services up Uses installer-managed stack context
memorymesh stop Stops managed Docker stack Pause local runtime Stack artifacts are preserved
memorymesh reset Stops stack and removes containers/volumes (based on flags) Clean stack state Confirm destructive flags carefully
memorymesh uninstall Stops stack and removes ~/.memorymesh Full local removal Destructive; use --yes for non-interactive
memorymesh upgrade Upgrade scaffold command Future upgrade path Placeholder/scaffold in current phase
memorymesh mcp Starts MCP stdio bridge Claude/Desktop or MCP client integration Long-running foreground process
memorymesh import:gpt --path Imports ChatGPT export JSON/ZIP paths Bring existing chat memory into MemoryMesh Requires runtime availability and valid source files
memorymesh import:documents --path Imports local documents (file or folder) Ingest notes, markdown, CSV/JSON datasets, and searchable project docs Recursive folder scan; unsupported files are skipped

Document Import

import:documents imports local files into MemoryMesh, recursively scans folders, parses supported formats, chunks text-based content where needed, and stores source metadata for source-aware search/retrieval.

Supported formats:

  • .txt
  • .md
  • .csv
  • .json
  • .jsonl
  • .ndjson

Run modes:

  • Interactive: run memorymesh, then choose Import documents
  • Direct command: memorymesh import:documents --path [options]

Detailed document-import limits, policies, project scoping, and checkpoint behavior are documented in packages/cli/README.md.

Claude MCP Integration

Setup can patch Claude Desktop config with:```json
{
"mcpServers": {
"memorymesh": {
"command": "memorymesh",
"args": ["mcp"]
}
}
}


Current integration behavior:

- Detects macOS and Windows Claude config paths
- Preserves unrelated config fields
- Handles missing config safely
- Fails safely on invalid JSON
- Shows restart guidance after successful update

## Diagnostics

`memorymesh doctor` includes checks for:

- Docker installed / daemon reachable
- MemoryMesh stack services (memorymesh, mongodb, neo4j, qdrant, ollama)
- MemoryMesh HTTP health
- Install/runtime config validity
- Embedding model presence
- Claude MCP config state

`memorymesh doctor --fix` can attempt safe repairs such as:

- Restarting stack
- Re-pulling embedding model
- Regenerating runtime env from install config
- Repairing Claude MCP config entry

## Troubleshooting

### Embedding model mismatch

This happens when existing vector data was created with a different embedding dimension (for example 768 vs 1024).

Resolution:

1. Run `memorymesh` to re-enter setup
2. Choose the target embedding mode (`flash` or `medium`)
3. Confirm reset when prompted if existing data must be rebuilt

Do not try to fix mismatch by manually exporting embedding environment variables.

1. Docker not installed
- Install Docker Desktop/Engine, then run `memorymesh doctor`

2. Docker daemon not running
- Start Docker Desktop/daemon, then run `memorymesh start`

3. Ollama model missing
- Run `memorymesh doctor --fix` or pull model via stack path, then re-run doctor

4. MemoryMesh health check fails
- Run `memorymesh doctor`; inspect stack status and restart with `memorymesh start`

5. Claude config missing
- This is non-destructive; create/open Claude Desktop once, then re-run integration

6. Claude config invalid JSON
- Fix JSON manually, then rerun installer or `memorymesh doctor --fix`

7. MCP integration not visible in Claude
- Confirm config contains `mcpServers.memorymesh`, then restart Claude Desktop

8. Bootstrap script cannot install prerequisites
- Use script guidance output to install Node/npm manually, then rerun script

9. `memorymesh` command not found after install
- Verify npm global bin is on `PATH`; reinstall with `npm install -g @okutucu/memorymesh`

10. npm install succeeds but runtime stack is unhealthy
- Run `memorymesh doctor` for root cause, then `memorymesh doctor --fix` for safe repairs and recheck

11. self-contained tarball/global install reports `EEXIST` binary conflict
- Remove conflicting global binary/package, or reinstall with a clean prefix, then run `npm install -g @okutucu/memorymesh`

## MemoryMesh Agent Usage

To enable MemoryMesh memory behavior in your AI agent, you must manually configure memory rules:

1. Go to:
   `Settings > Capabilities > Memory > Manage Memory`
2. Add TWO separate memory entries (this will use 2 memory slots out of the available slots, typically ~20).
3. Paste each of the following snippets into its own memory entry exactly as-is:```text
If MemoryMesh tools are available, use save_memory when: a decision is made (technical, personal, or workflow-related), something is learned or a bug is debugged, a preference or working style is clarified, or an important output is produced (code, email, document, plan, summary). Also save when the user says "remember this". Format every entry as: [INPUT] the user's request or question / [OUTPUT] the response, result, or artifact produced. No topic restriction.
``````text
If MemoryMesh tools are available, use search_memory proactively before generating a response if the topic might have relevant past context. Always search when: a new conversation starts and context feels missing, the user references something from a previous session ("last time", "we decided", "remember when"), or the current topic overlaps with decisions or outputs that may have been saved before. Trust the results — if a memory is retrieved, factor it into your answer.

Development / Monorepo Notes

Repository layout:

  • apps/server: server package
  • packages/core: shared types/utilities
  • packages/runtime: shared runtime/infrastructure
  • packages/cli: install/runtime CLI

Useful development commands:```bash
npm run build
npm test
npm run build -w @okutucu/memorymesh
npm test -w @okutucu/memorymesh


Developer source setup script (repo/dev workflow, not end-user install):```bash
./scripts/dev/install-dev.sh

Release/distribution notes (current state):

  • npm package target: @okutucu/memorymesh
  • bootstrap installer assets for end users are published as release assets (install.sh, install.ps1) from scripts/bootstrap/
  • release automation is scaffolded; treat it as evolving, not fully hands-off

常见问题

6 问
答案由本站依据条目字段整理,不引用上游原文
Memorymesh 48b5187b 怎么安装?

资料没有给出安装命令,请到官方仓库或文档按说明接入;这类条目多数是生态应用,或尚未发布可安装包。

Memorymesh 48b5187b 怎么在 Claude Desktop、Claude 里配置?

它的传输方式未标注,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Memorymesh 48b5187b 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Memorymesh 48b5187b 需要 API Key 吗?

资料标注需要会话登录认证,使用前准备好对应凭据,并按最小权限配置。

Memorymesh 48b5187b 支持哪些客户端?

资料标注支持 Claude Desktop、Claude;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Memorymesh 48b5187b 能提供哪些能力?

资料显示能力标签包括 AI工作流、本地优先、TypeScript、Claude、内存管理。具体工具清单以官方文档为准。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-17。