AI agents are useful, but they forget everything the moment a session ends. CMEM Cloud exists to remove the chore of re-explaining the same background every time. It is built by glass brain. The underlying engine is claude-mem, open source under Apache 2.0, which quietly records what was decided and where things got stuck as structured observations while the agent works. The paid cloud service syncs those records to the cloud and exposes them through a dedicated MCP link, so multiple agents such as Claude Code, Cursor, and Codex can connect to the same memory.
Key Features
- Automatic recording across sessions: Decisions, fixes, and dead ends that come up during an agent’s work are accumulated as “observations” automatically. You never have to stop and take notes yourself
- Offline-first cloud sync: The records live in a local database and are continuously mirrored to the cloud. Things keep working offline, and sync resumes once you are connected again
- Private MCP link: You get an endpoint scoped to your account. Paste it into Claude, your editor, or any MCP-aware agent and you can read and write the same memory. It is never shared, and it can be revoked or regenerated at any time
- Hybrid search: Full-text search (FTS5) is combined with semantic vector search and recency weighting to bring past context back
- Broad client support: Works with Claude Code, Cursor, Windsurf, Codex CLI, Gemini, OpenCode, and MCP-aware agents in general
- Privacy controls on an open-source base: There is a mechanism to exclude content you do not want recorded, and the engine itself is published on GitHub, so running it yourself without the cloud remains an option
Pricing
| Plan | Price | What’s included |
|---|---|---|
| claude-mem (open source) | Free (Apache 2.0) | Install the engine yourself. Cloud sync and the unified MCP link are not included |
| CMEM Cloud | $20/month | Cloud sync for the observations database |
| CMEM Pro | 7 days free, then $30/month | Memory generation runs on the provider’s observer model, cloud sync, private MCP link |
| Team Cloud | $333 per seat/month (3–50 seats) | Team use, onboarding with a forward deployed engineer |
Pricing is as of August 2026. The official site shows more than one figure in places, so check the official pricing page for the latest information.
Pros & Cons
✅ Pros
- No more repeating the same background every time, so long-running projects get moving faster
- Memory can be shared across multiple agents and machines, so context survives when you switch tools
- Recording happens automatically, with almost no operational overhead
- Because the engine is open source, you can try it for free before considering paid cloud sync
- Combining full-text and semantic search makes it easier to dig out past decisions you only half remember
⚠️ Cons
- At $20–$30 per month it is not cheap for an individual developer, and the only free option is self-hosting the open-source version
- Setup requires a runtime such as Node.js, so it is not something non-engineers can casually start with
- Work content accumulates as observation data, so confidential projects need exclusion settings and a look at the data retention policy
- It is a relatively new service, and features and pricing may keep changing
- The more records pile up, the more noise appears in search results, and deciding which context to recall is left to the user
Comparison with Similar Services
| Comparison | CMEM Cloud | Mem0 | Letta | Claude Projects |
|---|---|---|---|---|
| Main use | Working memory for coding agents | Memory layer embedded in apps | Framework for building agents with memory | Organizing conversation and document context |
| How records are made | Observations accumulate automatically during work | Developers store them via API | Managed inside the framework | Users register documents and instructions |
| Sharing across tools | Possible via the MCP link | Depends on your implementation | Depends on your implementation | Limited to Claude |
| Open source | Engine is Apache 2.0 | Partly public | Public | Not public |
| Typical users | Individual developers and dev teams | App developers | Agent developers | Claude users in general |
Who Is It For
- Developers who use Claude Code or Cursor daily and lose time repeating the same explanations
- People running several AI agents in parallel whose context is fragmented per tool
- Anyone on a long-running project who wants to trace back how a past decision was made
- People who want the same working memory across a local machine and a server, or home and office
- Those who want to start with the free open-source version and move to cloud sync when the need arises
Summary
CMEM Cloud addresses the basic limitation that AI agents forget, by automatically accumulating a work record and syncing it to the cloud. A helpful part of its design is the path it offers: try claude-mem for free, then move to the cloud version once you want the same memory available from multiple machines and agents. On the other hand, the price is not light for personal use, and confidential work calls for some thought about what gets recorded. Starting with the open-source version and seeing for yourself what persistent memory does to your workflow is a sensible first step.