Engineered by Dalian Fidwell Technology (fidwell.cn)
LongMemOS (MemoryBox 2.0) · v2.0.4

Local long-term memory OS shared across AI tools — trace answers back to exact source lines

Save conversations, codebase conventions, and debugging notes from Claude Code, Cursor, Codex, and ZeroOS workers on your own workstation or intranet server. Query from any connected client and LongMemOS retrieves original timestamped source passages before your chosen LLM responds — no built-in model weights, no re-indexing when switching models.

Brand update notice: MemoryBox 2.0 operates globally as LongMemOS (LongMemOS.com). Internal LongMemEval figures below reflect Fidwell testing on v2.0.4 without third-party audit.

Three Steps: Save -> Ask from Any Tool -> Verify Original Source

Illustrative sample data showing how cross-tool recall avoids vague summaries.

Cross-tool MCP memory trace · sample data LongMemOS.com
What You Do

Work inside Claude Code or Cursor on authentication fixes, schema constraints, or deployment incidents.

What LongMemOS Does

Ingests structured event records with timestamps, workspace tags, and lexical/semantic hybrid indexes locally.

Stored Output

Original decision wording stays intact in your local store — never overwritten by a lossy model summary.

memory_write({ scope: "repo/auth", decision: "Rotate JWKS keys every 24h; keep kid v1 during drain" })

Built for Engineering Teams, Not Chat Toy Storage

Separating durable retrieval from reasoning models keeps your memory fast, small, and portable.

Core Design

No Embedded Model Weights

embedded_models=false keeps the binary ~3.1 MB and decoupled from GPU VRAM. Upgrade your LLM anytime with zero database migration.

Available

Standard MCP + Local HTTP

Plug into Claude Code, Cursor, Codex, Hermes, and custom Python/TypeScript workflows through documented MCP endpoints.

Available

Cross-Language Retrieval

Search English codebase discussions using Japanese or Chinese phrasing — multilingual teams share one unified memory ledger.

Available

Exact Source Provenance

Every memory hit returns source document identifiers, timestamps, and verbatim passages so engineers verify facts immediately.

Available

Local & Appliance Sovereignty

Keep proprietary code rationale, client constraints, and incident post-mortems strictly inside your office LAN or Linux server.

Ecosystem

Native ZeroOS Memory Layer

Powers long-running multi-agent jobs inside ZeroOSAI.com so planners and executors reuse validated facts across runs.

Parent Orchestration Platform · ZeroOSAI.com

Need Full Multi-Step Task Scheduling Above Memory?

LongMemOS solves persistent context across tools. When you also want an AI coordinator that splits multi-step tasks, assigns Linux executors, tracks state in PostgreSQL, and retries failed steps automatically, pair LongMemOS with ZeroOS AI.

  • Self-hosted execution WebUI with step leases and idempotent replay.
  • Direct API routing to whichever reasoning models your organization already authorizes.
See ZeroOS AI Platform

Verification Matrix (v2.0.4)

Transparent reporting of measured dataset scores and physically verified hardware targets.

← Swipe horizontally to view full record →
Subject Version / Scope Recorded Outcome Conditions & Boundaries
LongMemEval Public Benchmark (500 Qs) MemoryBox / LongMemOS v2.0.4 463 / 500 (92.6%) Internal automated test run with embedded_models=false and configured external evaluator LLM; not independently verified by a third party.
Desktop & Server Binary Builds v2.0.4 release archive Win11 x86_64 · Linux x86_64 (~3.1 MB) Physically verified locally and on Linux servers (3.55–4.56 MiB unpacked on disk).
Edge ARM64 Appliance ARM64 RK3588 Linux Hardware-verified Tested on RK3588 industrial/AI box hardware for LAN-shared team memory; macOS desktop build currently under validation.

Editions & Delivery Formats

Deploy as lightweight workstation software, intranet container, or ready-to-plug hardware.

Individual / Team

Software Edition (Win11 / Linux)

Single lightweight binary + local data store for developers using Claude Code, Cursor, and Codex daily.

Request Evaluation Package
Team LAN

On-Premises Server Edition

Multi-workspace deployment inside company servers so engineering, QA, and support agents query unified memory.

Discuss Team Licensing
Turnkey Hardware

ARM64 RK3588 Memory Box

Pre-installed quiet LAN appliance built around verified RK3588 hardware — plug power and Ethernet in your office.

Check Hardware Availability
Technical Reference: MCP Integration & Binary Specs (expand to read)

Package & Runtime Specifications

  • Compressed archive size: ~3.1 MB; unpacked runtime footprint: 3.55–4.56 MiB depending on OS architecture.
  • Configuration flag: embedded_models=false by default — zero hidden model weights bundled.
  • Supported transport interfaces: Model Context Protocol (MCP) + REST HTTP endpoints for custom agent pipelines.

Scope Clarification on Benchmarks

  • No cross-vendor head-to-head ranking claims are published where identical hardware/prompt test conditions were not controlled.
  • LongMemEval score (463/500) represents Fidwell internal test execution on v2.0.4 over the public 500-question benchmark.

Frequently Asked Questions

Why is the product branded as LongMemOS overseas when some docs mention MemoryBox?

MemoryBox 2.0 is the hardware/appliance lineage in China. Globally, we use LongMemOS (LongMemOS.com) to clearly signify a dedicated Long-Term Memory OS for AI agents and avoid naming confusion with unrelated consumer photo-keepsake websites.

Does LongMemOS send my code or prompts to Fidwell servers?

No. LongMemOS runs locally on your PC, office LAN appliance, or company Linux server. Your stored memory remains on storage you control.

What happens if I switch from Claude to GPT or a local model later?

Your stored history, indexes, and MCP interface stay untouched because LongMemOS stores structured records independently of any LLM vendor.

Is macOS supported today?

Win11 x86_64, Linux x86_64, and ARM64 RK3588 are verified. macOS support is currently undergoing validation — contact us to join validation testing.

Book an Engineering Walkthrough

Share how your team runs coding tools or internal automation today. Only your name and one contact method are required.

Direct Contact · Dalian Fidwell Technology

All consultation records go directly to Fidwell engineering support. We respond during standard business hours (UTC+8).

Email: fidwellservice@fidwell.cn

Tel: +86-411-81981709

Address: Room 1303, Bank of East Asia Tower, No. 7 Renmin Road, Zhongshan District, Dalian, Liaoning, China

Send Inquiry

Information submitted here is stored only for responding to your demo or evaluation request and is never sold or shared with unrelated third parties.