📄 工具详情 ⚙️ 安装教程 📚 使用教程
能力标签
🔌 MCP🤖 Agent🔄 工作流💻 CLI🔗 REST API🧬 Embedding🧠 Claude✨ GPT⌨️ Cursor🖥 本地 LLM
非线性内存OS
🔌
MCP工具

非线性内存OS

基于 TypeScript · 让 AI 助手直接操作你的系统与工具
英文名：nlm-memory
⭐ 6 Stars 💻 TypeScript 📄 Apache-2.0 🏷 AI 7.5分
7.5AI 综合评分
aiaiderclaude-codecodexcursortypescript
⬇ 下载源码 ZIP ⚙️ 配置说明 📺 TG 频道
✦ AI Skill Hub 推荐

非线性内存OS 是 AI Skill Hub 本期精选MCP工具之一。综合评分 7.5 分，整体质量较高。我们推荐使用将其纳入你的 AI 工具库，帮助提升工作效率。

📚 深度解析

非线性内存OS 是一款基于 MCP（Model Context Protocol）标准协议的 AI 工具扩展。MCP 协议由 Anthropic 开发并开源，旨在建立 AI 模型与外部工具之间的标准化通信接口，目前已被 Claude Desktop、Claude Code、Cursor 等主流 AI 工具采纳。

通过安装 非线性内存OS，你的 AI 助手将获得额外的工具调用能力，可以用自然语言直接操控该工具的功能，无需学习复杂的命令行语法。MCP 工具的核心价值在于"一次配置，永久增强"——配置完成后，每次与 AI 对话时都可以无缝调用这些工具。

在技术实现上，MCP 工具通过标准的 JSON-RPC 协议与 AI 客户端通信，工具的功能以"工具列表"的形式暴露给 AI 模型，AI 可以按需调用。非线性内存OS 提供了结构化的工具调用接口，使 AI 模型能够精确地理解和使用每个功能点，显著降低 AI 在工具使用上的错误率。

与传统的 API 集成相比，MCP 工具的优势在于无需编写代码——用户只需在配置文件中添加几行 JSON，即可让 AI 获得全新能力。AI Skill Hub 将 非线性内存OS 评为 AI 评分 7.5 分，属于同类工具中的优质选择。

📋 工具概览

非线性内存OS 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。

GitHub Stars
⭐ 6
开发语言
TypeScript
支持平台
Windows / macOS / Linux
维护状态
轻量级项目，按需更新
开源协议
Apache-2.0
AI 综合评分
7.5 分
工具类型
MCP工具
Forks
—

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

非线性内存OS 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。

📌 核心特色
  • 通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
  • 提供结构化工具调用接口，显著降低 AI 集成复杂度
  • 支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
  • 可与其他 MCP 工具组合叠加，构建完整 AI 工作站
  • 轻量无侵入设计，不影响现有系统架构
🎯 主要使用场景
  • 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
  • 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
  • 将多个 MCP 工具组合使用，构建个人专属 AI 工作站
以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。
安装命令
# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/pbmagnet4/nlm-memory

# 方式二：手动配置 claude_desktop_config.json
{
  "mcpServers": {
    "-----os": {
      "command": "npx",
      "args": ["-y", "nlm-memory"]
    }
  }
}

# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
📋 安装步骤说明
  1. 确认已安装 Node.js（v18 或以上版本）
  2. 打开 Claude Desktop 或 Claude Code 的 MCP 配置文件
  3. 按「交给 Agent 安装 → Claude Desktop」标签中的 JSON 配置填入 mcpServers 字段
  4. 保存配置文件并重启 Claude 客户端
  5. 重启后，在对话中即可使用本工具
以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。
常用命令 / 代码示例
# 安装后在 Claude 对话中直接使用
# 示例：
用户: 请帮我用 非线性内存OS 执行以下任务...
Claude: [自动调用 非线性内存OS MCP 工具处理请求]

# 查看可用工具列表
# 在 Claude 中输入："列出所有可用的 MCP 工具"
以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。
配置示例
// claude_desktop_config.json 配置示例
{
  "mcpServers": {
    "_____os": {
      "command": "npx",
      "args": ["-y", "nlm-memory"],
      "env": {
        // "API_KEY": "your-api-key-here"
      }
    }
  }
}

// 保存后重启 Claude Desktop 生效
📑 README 深度解析 真实文档 完整度 81/100 查看 GitHub 原文 →
以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

简介

<p align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="assets/banner-dark.svg" /> <source media="(prefers-color-scheme: light)" srcset="assets/banner-light.svg" /> <img alt="nlm-memory — Local-first non-linear memory OS" src="assets/banner-light.svg" width="720" /> </picture> </p>

<p align="center"> <a href="https://www.npmjs.com/package/nlm-memory"><img src="https://img.shields.io/npm/v/nlm-memory?color=CB3837&label=npm&logo=npm" alt="npm version" /></a> <a href="https://github.com/pbmagnet4/nlm-memory/actions/workflows/ci.yml"><img src="https://github.com/pbmagnet4/nlm-memory/actions/workflows/ci.yml/badge.svg?branch=main" alt="CI status" /></a> <a href="https://github.com/pbmagnet4/nlm-memory/blob/main/LICENSE"><img src="https://img.shields.io/github/license/pbmagnet4/nlm-memory?color=blue" alt="License: Apache 2.0" /></a> <a href="https://nodejs.org"><img src="https://img.shields.io/node/v/nlm-memory?color=brightgreen" alt="Node 20+" /></a> <img src="https://img.shields.io/badge/tests-2%2C000%2B%20passing-success" alt="2,000+ tests passing" /> <img src="https://img.shields.io/badge/MCP-9%20runtimes-8A2BE2" alt="MCP across 9 runtimes" /> <img src="https://img.shields.io/badge/hooks-4%20runtimes-7B2CBF" alt="Hooks on 4 runtimes" /> <img src="https://img.shields.io/badge/telemetry-none-informational" alt="Zero telemetry" /> </p>

<p align="center"> <a href="#install">Install</a> &middot; <a href="#quick-start">Quick Start</a> &middot; <a href="#runtimes">Runtimes</a> &middot; <a href="#how-recall-works">How recall works</a> &middot; <a href="#agent-self-improvement-signals">Signals</a> &middot; <a href="#mcp-tools">MCP</a> &middot; <a href="#rest-api">REST API</a> &middot; <a href="#daily-digest">Digest</a> &middot; <a href="#configuration">Config</a> &middot; <a href="#security">Security</a> &middot; <a href="#vs-alternatives">vs Alternatives</a> </p>

---

nlm-memory is a local-first memory layer for AI coding agents. It indexes every session from Claude Code, Codex CLI, OpenCode, Cursor, Windsurf, Hermes WebUI, Hermes Agent, Aider, and pi across nine runtimes into a single searchable store on your machine. Three properties no other memory layer ships together:

  1. Cross-runtime reach. One index, every adapter.
  2. Editable timeline. Sessions can be superseded by newer ones; entities can be retired. Patch history retroactively — no other tool lets you do this. See docs/supersedence.md.
  3. Strong R@5 baselines. Two measurements, both committed: 97.2% R@5 on the public LongMemEval-S benchmark (hybrid mode, n=500, body-only, reproducible by anyone) and 90.0% R@5 on a real 14-month personal corpus (n=40 curated decision queries, keyword mode, 36/40 retrieved). No fine-tuning. The corpus labels were generated by DeepSeek V4; expect a lower number with a smaller local classifier. See docs/benchmarks.md and docs/methodology-recall-baseline.md.

Install-and-recovery validation. A context-free agent ("the stranger") with no knowledge of NLM internals installed from HEAD and answered "Why pgvector over Qdrant, and was anything reconsidered?" correctly in under 6 minutes, recovering the full decision arc including a superseded session. See reports/stranger-sim/2026-06-10-recovery-simulation.md.

Everything stays on your machine by default. No telemetry, no account. The classifier defaults to local (Ollama); if you opt into a cloud classifier (DeepSeek, OpenAI, Anthropic, OpenRouter, or any OpenAI-compatible endpoint), session transcripts are sent to that provider — see Security for the exact data-flow.

---

What's inside the UI

Open http://localhost:3940/ui after the daemon starts.

PageWhat it shows
**Live**Sessions being written in real time, recent reads, recent decisions
**Pulse**System health — coherence, runtimes, stale entities, recent sessions
**River**Full session timeline with density controls + superseded-lane visualization
**Thread**Per-entity conversation history with runtime filters and ←/→ navigation
**Search**Keyword, semantic, or hybrid recall with match snippets and field-origin tags
**Recall**Adoption telemetry — useful_hit_rate, source breakdown, query log
**Settings**Sources, providers, classifier, data backup/restore

---

Install

npm install -g nlm-memory
nlm setup

nlm setup is the interactive first-run wizard. It picks your classifier + model, wires the runtimes you actually use, generates an NLM_MCP_TOKEN, hardens permissions on ~/.nlm/, and installs the daemon supervisor for your platform. For headless environments (servers, CI, agents), use nlm start directly after initial setup — it boots the daemon without interactive prompts.

PlatformDaemonNotes
**macOS**LaunchAgent at ~/Library/LaunchAgents/com.github.pbmagnet4.nlm-memory.plistAuto-starts on login
**Linux**systemd user unit at ~/.config/systemd/user/nlm.serviceHeadless servers: loginctl enable-linger $USER so the daemon survives logout
**Windows**Manual nlm start for nowHook + MCP install paths are platform-aware; supervisor lands next release

Stop or remove: nlm uninstall.

---

Quick Start

After nlm setup finishes, open http://localhost:3940/ui — the daemon is running. (If you changed NLM_PORT, substitute your port in all examples below.) A 30-second sanity check:

nlm recall "what was that pgvector decision"   # one-shot search from the shell
nlm digest                                      # yesterday's activity at a glance
nlm --version

---

Configuration and privacy

VarDefaultWhat
NLM_SIGNALS_ENABLED1 (on)Set to 0 to disable signal ingest entirely
NLM_SIGNAL_RETENTION_DAYS90Raw signals older than this are pruned

Signals are local-only. They are stamped with a per-install ID from ~/.nlm/install-id and never leave the machine.

Configuration

Environment variables

VarDefaultWhat
NLM_PORT3940Daemon bind port (loopback only)
NLM_DB_PATH~/.nlm/canonical.sqliteSQLite canonical store location
NLM_HOOK_MODElivelive injects pointer block; shadow logs without injecting
NLM_HOOK_RECALL_TIMEOUT_MS4000Hook-side budget for the daemon recall call. Must stay below NLM_HOOK_DEADLINE_MS and above the daemon's embed deadline or the pointer block silently empties (#396).
NLM_HOOK_DEADLINE_MS6000Outer wall-clock deadline for the recall + gate stages of a hook fire. Fail-open on expiry.
NLM_HOOK_PROMPT_RECALL(unset = off)Per-prompt ambient recall. Off by default (pull-first); set on to inject a pointer block on every prompt. off disables explicitly.
NLM_HOOK_LOG~/.nlm/hook-log.jsonlHook fire log; powers digest's liveness alert
NLM_USEFUL_HIT_LOG~/.nlm/useful-hit-log.jsonlCitation/useful-hit ledger
NLM_QUERY_LOG~/.nlm/query-log.jsonlRecall query telemetry
NLM_CITATION_LOG~/.nlm/citation-log.jsonlStop-hook citation events
NLM_MISS_LOG~/.nlm/miss-log.jsonlStop-hook miss events — sessions the agent explicitly fetched via get_session/cite_session that the hook never surfaced. Reviewed via nlm misses.
NLM_MISS_LOG_ENABLEDtrueSet to 0 to disable miss-log emission entirely.
NLM_MCP_TOKENauto-generated256-bit bearer for /api/* (non-browser) and /mcp
NLM_MCP_CONFIG~/.mcp.jsonPath the connect/disconnect commands modify
NLM_CLASSIFIERollamaollama (local, default), deepseek, openai, anthropic, openrouter, or openai-compatible
NLM_CLASSIFIER_MODELqwen3.5:4bModel id for the chosen provider. qwen3.5:4b (Q4_K_M, 4.7B) is the recommended local default. For measured tier expectations, see [docs/classifier-tiers.md](docs/classifier-tiers.md).
NLM_OLLAMA_URLhttp://localhost:11434Override Ollama endpoint
NLM_CODE_EXEMPLARS_ENABLED—Set to 1 to enable the code-exemplar lane (capture from code-bearing signals + the recall_code tool/routes). Off by default.
NLM_CODE_EMBED_MODELhf.co/awhiteside/CodeRankEmbed-Q8_0-GGUFOllama model for code-exemplar embeddings (768-dim). Only used when the exemplar lane is enabled.
NLM_ADAPTERSallComma-separated allowlist of adapters to enable
DEEPSEEK_API_KEY—Required only when classifier=deepseek
NLM_DISABLE_UPDATE_CHECK—Set to 1 to disable the daily npm-registry update check
NLM_RECALL_DECAY_HALF_LIFE_DAYS180Half-life of the recency multiplier applied to recall scores. Older sessions score lower; defaults to 6 months. Set to 0 to disable recency weighting entirely.
NLM_RECALL_DECAY_FLOOR0.25Lower bound on the recency multiplier — even ancient sessions retain at least 25% of their raw score so a perfect-match old session can still surface.
NLM_RECALL_REWRITE_DEFAULTtrueDefault value for the MCP recall_sessions rewrite parameter. When true, the service runs an LLM rewrite on vague natural-language queries before search. The HTTP hook caller bypasses rewrite regardless (hot-path protection).
NLM_RECALL_REWRITE_TIMEOUT_MS5000Per-call timeout for the rewrite LLM. Separate from the classifier timeout.
NLM_FACT_CORROBORATION_BOOST_CAP2.0Maximum multiplicative boost applied to fact recall scores based on how many sessions corroborate the same (subject, predicate, value). Log-scale: 1 corroboration is 1.0×, 10 is 2.0× (capped). Set to 1.0 to disable the boost — the count is still returned on each hit.
NLM_HOOK_INJECT_FACTStrueWhether to attach high-confidence facts about top-hit entities to the pointer block injected by the hook. Set to 0 to disable globally.
NLM_HOOK_FACT_LIMIT5Maximum number of facts in the "Known facts" section of the pointer block.
NLM_HOOK_FACT_MIN_CORROBORATION2Minimum number of sessions that must have asserted a fact before it qualifies for hook injection. Set to 1 to include single-source facts.
NLM_HOOK_FACT_MIN_CONFIDENCE0.7Minimum classifier confidence for a fact to qualify for hook injection.
TELEGRAM_BOT_TOKEN / TELEGRAM_CHAT_ID—Required for nlm digest --telegram
NLM_ALERT_WEBHOOK—Set to a URL to enable daemon self-reporting: version-drift and embedder-cold transitions POST a CloudEvents-shaped payload here. Unset means zero network calls.
NLM_ALERT_WEBHOOK_TOKEN—Optional bearer token sent as Authorization: Bearer <token> on every NLM_ALERT_WEBHOOK POST. Only used when the webhook is set.
NLM_JOB_STALL_MINUTES20Minutes a daemon-supervised reprocess run (see POST /api/jobs/reprocess) can go without a progress advance before the supervisor calls it stalled and fires nlm.job.stalled via NLM_ALERT_WEBHOOK. The same event type also fires with reason: "exhausted" (restart budget used up) and reason: "spawn_failed" (the child process itself never launched — EMFILE/ENOMEM/bad binary).
NLM_WIKI_DIR— (opt-in, off by default)The wiki projection is off unless this is set. When set, every NLM_WIKI_INTERVAL_HOURS the SQLite-backend daemon rolls up fact subjects that cleared the page threshold into a markdown tree at <dir>/Memory, reconciled against the previous run so an unchanged page writes zero bytes. Unset means no timer, no directory, no log line — an upgrade never surprises you with new files under your home directory. A failed run leaves the previous tree untouched and fires nlm.wiki.projection_failed via NLM_ALERT_WEBHOOK; a run whose written page count doesn't match what the corpus currently qualifies for fires nlm.wiki.coverage_drift. Postgres backend: not implemented, daemon never starts it.
NLM_WIKI_INTERVAL_HOURS6How often the wiki projection re-runs once NLM_WIKI_DIR is set.
NLM_WIKI_MIN_FACTS3Minimum current (non-superseded, non-retired) fact count a subject needs to earn a wiki page.
NLM_WIKI_MIN_SESSIONS3Minimum distinct contributing sessions a subject needs to earn a wiki page.
NLM_WIKI_LINK_BASEhttp://127.0.0.1:<NLM_PORT>Base URL a wiki page's session links point at. Override when the vault is synced somewhere the daemon's own loopback address won't resolve.

Config file

~/.nlm/.env — autoloaded by every CLI command. Mode 0600, owned by you, never readable by other users. The setup wizard writes the initial keys; you can edit it directly.

Reference producer

The pi quality-gate extension (in the pi-sandbox repo) is a ~10-line integration that emits nlm.signal per gate step and again on retry exhaustion. It is the canonical example of the session-embedded transport.

REST API

Daemon binds 127.0.0.1:3940 (override with NLM_PORT). Selected endpoints:

MethodPathAuthPurpose
GET/api/healthHost-onlyLiveness probe: version, warmup + embedding-lane state, corpus stats, update: {current, latest, behind} from the daily npm check, and job: {name, state, processed, total, startedAt, lastAdvanceAt, restarts, restartsTotal} \| null — the daemon-supervised reprocess run (null when idle: never started, or the daemon just restarted, since the supervisor holds no state across restarts). restarts is the restarts-without-progress counter (resets on any advance); restartsTotal is the run's lifetime respawn count, never reset — it's the one to watch for OOM churn on a run that keeps making progress between crashes.
GET/api/recallBearer/OriginHybrid recall — ?q=, ?mode=keyword\|semantic\|hybrid, ?limit=, ?include_superseded=true (default off; opt-in to surface overturned sessions down-ranked, badged with superseded_by). Each result carries status + superseded_by.
GET/api/recall/statsBearer/Origin7-day stats: total, hit_rate, useful_hit_rate, top queries
GET/api/recall/recentBearer/OriginLast N recall events for live tail/telemetry
GET/api/recall/cite-statsBearer/OriginCitation rate over ?days=
GET/api/session/:idBearer/OriginFull session body + supersedence links
GET/api/recall/factsBearer/OriginStructured fact search; includes hint field when empty
GET/api/facts/historyBearer/OriginVersion chain for one fact
GET/api/datasetBearer/OriginFull session list for the UI dataset view
GET/api/live/recent-writesBearer/OriginLive tail of ingested sessions
GET/api/data/backupBearer/OriginStreaming SQLite snapshot download
POST/api/data/restoreBearer/OriginStage a snapshot for apply-on-restart
POST/api/citation/explicitBearer/OriginLog an explicit session citation (the route backing the cite_session MCP tool; payload {id, conversation_id?, reason?})
POST/api/hook/pre-compactBearer/OriginHook endpoint; flushes the surfaced-IDs memo
ALL/mcpBearer requiredStreamable-HTTP MCP transport — the default for every runtime; stateless, team resolved from the token
GET/api/recall-codeBearer/OriginSemantic code-exemplar search — ?q=, ?repo=, ?lang=, ?k=, ?negatives=0. Requires NLM_CODE_EXEMPLARS_ENABLED=1.
POST/api/exemplarBearer/OriginDirect code-exemplar push (capture also happens automatically from code-bearing signals). Requires the flag.
POST/api/jobs/reprocessBearer/OriginStart a daemon-supervised reprocess run (body: optional {args: string[]}, forwarded verbatim as CLI flags). 202 + snapshot on start, 409 + snapshot if one's already active. Backs nlm reprocess --daemon.
DELETE/api/jobs/reprocessBearer/OriginStop the active supervised run (no-op if idle/already terminal). 200 + snapshot.

/api/* is gated by three layers: 127.0.0.1 Host check (defeats DNS rebinding), Origin check when the browser sends one (defeats cross-origin drive-by), Bearer fallback when Origin is absent (server-to-server clients).

---

Pipeline

What happens when an AI runtime writes a session and you later recall it:

ingest:  runtime transcript (jsonl/sqlite)
   -> adapter parses runtime-specific format
   -> classifier (Ollama local by default; DeepSeek / OpenAI / Anthropic / OpenRouter / OpenAI-compatible if you opt in) extracts label + entities + decisions + open questions
   -> embedder (nomic-embed-text via Ollama) computes 768-dim vector
   -> SQLite canonical store + FTS5 keyword index + sqlite-vec ANN index

recall: prompt / query
   -> tokenize + match scoring (label x3, entity-exact x4, decision x2, summary x1, phrase-bonus +5)
   -> hybrid: BM25-style keyword + vector cosine, fused by score
   -> recency decay (180-day half-life, floored) x supersedence down-rank
   -> select-top-N gate (per-fire cap 3, per-conversation cap 10)
   -> pointer block prepended to model context (hooks) or returned as tool result (MCP)

---

vs Alternatives

**nlm-memory**mem0Letta / MemGPTBuilt-in (CLAUDE.md)
**Unit of memory**Whole session + extracted markersAtomic factsGraph nodes + edgesStatic file
**Cross-runtime**9 adapters, one corpusPer-app SDK integrationPer-app SDK integrationPer-runtime config
**Editable timeline**Sessions can be superseded, retired, abortedAppend-only fact logGraph editsManual file edits
**R@5 baseline**97.2% public benchmark; 90.0% real 14mo corpus ([docs/benchmarks.md](docs/benchmarks.md))published variespublished variesn/a
**External deps**SQLite + Ollama (local)Postgres or QdrantPostgresnone
**Hosted offering**none — local onlyyesyesn/a
**Account required**noneyes (cloud tier)yesnone
**Telemetry**noneyesyesnone
**License**Apache 2.0Apache 2.0Apache 2.0—

The defining property is the editable timeline. mem0 and Letta append; NLM lets you reach back and mark a session as superseded by a newer one, retire one as no-longer-relevant, or flag one as aborted-mid-flight. The next recall surfaces the corrected version, not the stale one. A claim from 6 months ago can be patched today.

---

🇨🇳 中文文档镜像 AI 翻译 2026-06-12
英文原文章节由系统翻译为中文摘要，便于快速理解。完整原文见上方 "📑 README 深度解析"。
📌 简介

本项目是 nlm-memory 的中文文档，介绍了一个本地化的非线性内存操作系统。项目的 logo 和 banner 可以在 assets 文件夹中找到。

⚡ 功能介绍

UI 页面包括 Live、Pulse、River 和 Thread 四个部分，分别显示会话的实时写入、系统健康、会话时间线和实体对话历史。

🛠 安装步骤（Docker/pip/源码）

安装步骤包括使用 npm 安装 nlm-memory 和 nlm setup，后者是一个交互式的首次运行向导，用于选择分类器和模型，配置运行时，生成 MCP token，并安装守护进程。

🚀 使用教程

使用教程包括启动守护进程，打开 UI 页面，使用 nlm recall 和 nlm digest 等命令进行搜索和查看活动。

⚙️ 配置说明（含 MCP / env）

配置说明包括环境变量和关键参数，例如 NLM_SIGNALS_ENABLED、NLM_SIGNAL_RETENTION_DAYS、NLM_PORT 等。

🔌 API 说明

API 参考文档包括 REST API 端点，例如 /api/health 和 /api/recall，用于检查守护进程的健康状态和进行混合回忆。

🔄 工作流/模块

工作流和模块说明包括 AI 运行时写入会话和用户回忆会话的过程，涉及 ingest、adapter、分类器和回忆等步骤。

🎯 aiskill88 AI 点评 A 级 2026-06-12

高质量的MCP工具，为AI运算提供了便捷的内存管理

📚 实用指南（长尾问题）
适合谁
  • 使用 Cursor 编辑器、希望提升 AI 编程效率的开发者
  • 需要让 Claude / Cursor 操作本地工具的 AI 工程师
  • 构建多智能体协作系统的 Agent 开发者
  • 构建企业知识库 / RAG 检索应用的团队
最佳实践
  • 配置 MCP 服务器时建议使用 stdio 传输 + JSON-RPC，避免暴露公网
  • 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
  • Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
  • Cursor rules 控制在 80 行内，否则模型上下文成本会显著上升
常见错误
  • API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
  • MCP 配置路径拼错或权限不足，重启 Claude Desktop 才生效
  • 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
部署方案
  • CLI：直接 npm install -g / pip install，命令行调用
  • 本地部署：CPU 8GB 起，GPU 推荐 16GB+ 显存
  • 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
相关搜索
nlm-memory 中文教程nlm-memory 安装报错怎么办nlm-memory MCP 配置nlm-memory Agent 工作流nlm-memory 与同类工具对比nlm-memory 最佳实践nlm-memory 适合谁用

⚡ 核心功能

👥 适合谁
  • 使用 Cursor 编辑器、希望提升 AI 编程效率的开发者
  • 需要让 Claude / Cursor 操作本地工具的 AI 工程师
  • 构建多智能体协作系统的 Agent 开发者
  • 构建企业知识库 / RAG 检索应用的团队
⭐ 最佳实践
  • 配置 MCP 服务器时建议使用 stdio 传输 + JSON-RPC，避免暴露公网
  • 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
  • Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
  • Cursor rules 控制在 80 行内，否则模型上下文成本会显著上升
⚠️ 常见错误
  • API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
  • MCP 配置路径拼错或权限不足，重启 Claude Desktop 才生效
  • 显存不足直接 OOM — 优先降低 context 或换更小的量化模型

👥 适合人群

Claude Desktop / Claude Code 用户AI 工具开发者需要扩展 AI 能力的专业人士自动化工程师

🎯 使用场景

  • 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
  • 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
  • 将多个 MCP 工具组合使用，构建个人专属 AI 工作站

⚖️ 优点与不足

✅ 优点
  • +Apache-2.0 协议，可免费商用
  • +标准化 MCP 协议，生态互联性强
  • +与 Claude 官方生态无缝对接
  • +即插即用，配置简单快捷
⚠️ 不足
  • −依赖 Claude 客户端，非 Claude 用户无法使用
  • −MCP 协议仍在持续演进，接口可能变更
  • −需要一定的配置步骤
⚠️ 使用须知

AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。

建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。

📄 License 说明

✅ Apache 2.0 — 宽松开源协议，可商用，需保留版权声明和 NOTICE 文件，含专利授权条款。

🔗 相关工具推荐

mcp-use MCP工具
MCP工具
私有GPT
MCP工具
playwright-mcp MCP工具
基于Playwright的MCP协议服务器实现，提供浏览器自动化能力。支持网页交互、截图、数据采集等功能，适合开发者集成
FastGPT
基于LLM的知识平台，提供综合解决方案
📰 相关 AI 新闻
AI 前沿资讯：We made 8 AIs bet on the FIFA …
AI 资讯 · 知识关联
AI Agent 自主化能力最新进展
AI 资讯 · 知识关联
AI Agent 自主化能力最新进展
AI 资讯 · 知识关联
OpenAI 最新产品与研究动态
AI 资讯 · 知识关联
🍿 AI 圈相关吃瓜
配了5个 MCP 工具，Claude 一个都没用
AI 圈观察
AutoGPT 自主完成了任务：把我的文件夹全部重命名了
AI 圈观察
Filesystem MCP 帮 Claude 找文件，找了整个 node_modules
AI 圈观察
🗺️ 相关解决方案
mcp
mcp-server-setup
agent
ai-agent-workflow
workflow
ai-workflow-templates
cli
cli-productivity
🧩 你可能还需要
基于当前 Skill 的能力图谱，自动补全的工具组合
gptme Agent工作流
MCP · Agent · 工作流
mcp-proxy-for-aws MCP工具
为AWS服务提供MCP（模型上下文协议）代理的开源工具。支持SigV4签名认证，让大模型能通过标准MCP接口调用AWS API，适合需要集成
Augustus
LLM安全测试框架，检测prompt注入、越狱等
客户支持聊天机器人
基于检索增强的全生产就绪AI客户支持系统
MassGen多智能体系统
MCP · Agent · 工作流
代码智能引擎
MCP · Agent · 工作流

❓ 常见问题 FAQ

nlm-memory 是一款TypeScript开发的AI辅助工具。开源MCP工具：Local-first non-linear memory OS for AI operators. One index across Claude Code,。⭐6 · TypeScript 主要应用场景包括：AI运算和数据处理。
💡 AI Skill Hub 点评

经综合评估，非线性内存OS 在MCP工具赛道中表现稳健，质量良好。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。

⬇️ 获取与下载
⬇ 下载源码 ZIP

✅ Apache-2.0 协议 · 可免费商用 · 直接从 aiskill88 服务器下载，无需跳转 GitHub

📚 深入学习 非线性内存OS
查看分步骤安装教程和完整使用指南，快速上手这款工具
⚙️ 安装教程 📚 使用教程
🌐 原始信息
原始名称 nlm-memory
原始描述 开源MCP工具：Local-first non-linear memory OS for AI operators. One index across Claude Code,。⭐6 · TypeScript
Topics aiaiderclaude-codecodexcursortypescript
GitHub https://github.com/pbmagnet4/nlm-memory
License Apache-2.0
语言 TypeScript
🔗 原始来源
🐙 GitHub 仓库  https://github.com/pbmagnet4/nlm-memory 🌐 官方网站  https://www.npmjs.com/package/nlm-memory

收录时间：2026-06-12 · 更新时间：2026-06-13 · License：Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。

📺 订阅 AI Skill Hub Daily Telegram 频道
每天 8 条精选 AI Skill、MCP、Agent 与自动化工具推送
加入频道 →