Compare/Claude
ContxOS vs Claude Memory, Projects and Managed Agents Memory
Claude now has memory in its apps, per-project memory, a memory tool in the API and memory stores for Managed Agents. Here is how each compares with a ContxOS knowledge base.
Jose R. Gonzalez, CPA · Founder of ContxOS
Updated · 3 min read
In short
Claude’s memory remembers you: editable topic files shared by Claude chat and Cowork, isolated per Claude Project. Claude Managed Agents add versioned memory stores and a Dreaming preview for developers. ContxOS remembers the work: typed decisions, handoffs and workflows that Claude.ai, Claude Code and any other MCP client share. They are complementary, and ContxOS runs inside Claude as a connector.
Side by side
| Capability | ContxOS | Claude app memory | Managed Agents memory |
|---|---|---|---|
| What it is | A hosted knowledge base that AI agents read and write while they work on your projects. | Memory of you across Claude chats and Cowork, stored as topic files. | Workspace memory stores that developers mount into Managed Agents sessions. |
| Where it works | Any MCP client with OAuth sign-in (Claude.ai, Claude Code and others). One memory, shared by every connected agent. | Claude chat and Cowork. The Aug 2026 post does not extend it to Claude Code, whose auto memory is machine-local. | Claude Managed Agents sessions (API). |
| How memory is written | Deliberately: when a session closes, in a handoff, when you say "record this", or as a workflow step. No transcript ingestion. | Automatically, as individual topics while you chat. You can view, edit, add or delete. | The agent edits text files with file tools. Every change creates an immutable version. |
| Structure | Typed entries (decision, result, plan, TODO, bug, finding, note…) joined by typed links (supersedes, resolves, implements, supports, refutes…), each with an authority level. | Topic files. Typed entries and links are not documented. | Text documents in a directory (up to 100 kB each, 10,000 per store). |
| Keeping memory current | Every search result shows its date and whether it is superseded (with a pointer to the replacement), expired or in conflict. Weak matches are labeled low confidence. | Topics edited in place. | Version history (kept 30 days) and optional concurrency checks. Dreaming (research preview) writes a reorganized store with contradictions replaced. |
| Session continuity | Structured handoffs (accomplishments, pending work by priority, blockers by severity, decisions). "resume" restores one in a single step. | Chat search (paid plans) finds and cites past chats. | Not documented beyond memory stores. |
| Procedures | Workflows (stages, steps, per-client variants). Runs track position, verified facts and holds outside the chat, and proven steps can go to subagents. | Separate feature: Skills. | Separate feature: Skills (out of beta Aug 2026). |
| Separation and sharing | Projects, optional buckets within a project, per-project members, seats. | Each Claude Project has isolated memory. Team plans can share Projects. | Workspace-scoped stores, up to 8 per session, read-write or read-only. |
| Price | Free plan with one project and one seat. Paid plans add projects and seats (see contxos.com/pricing). | Included in Claude plans, Free included. Admin-controlled on Team and Enterprise. | $0.08 per session-hour plus tokens. No separate memory price listed. |
Competitor details come from their own documentation as of September 29, 2026 (see Sources). Products change quickly; tell us if something here is out of date.
What changed in Claude’s memory this year
Claude’s app memory moved to individual, categorized entries in July 2026. In August, chat and Cowork began sharing one memory, visible as files under Topics that you can read, edit or delete. Sensitive topics are off by default. Memory is on by default for Free, Pro and Max. On Team and Enterprise, admins control it.
On the developer side, memory stores for Claude Managed Agents entered public beta on April 23, 2026, with immutable versioning. Dreaming, a research preview announced May 19, 2026, reviews up to 100 past sessions and writes a cleaned-up copy of a store for review.
The difference in one example
Say you decide to switch a project from monthly to annual billing. Claude’s memory might note that you prefer annual billing. ContxOS records a decision with its rationale, links it as superseding the earlier monthly-billing decision, and attaches it to the session where it was made. Next month, an agent in Claude Code that searches for billing finds the annual decision, sees that the monthly one is superseded, and knows why.
Using them together
ContxOS is added to Claude as a custom connector and runs inside Claude.ai and Claude Code. Leave Claude’s memory on for personal context, and let ContxOS carry the project record. Anthropic’s own guidance for multi-session agent work (a progress log plus a checklist) is the same pattern ContxOS handoffs and workflows provide, ready-made and shared across clients.
Which one fits
ContxOS fits when
- You work in both Claude.ai and Claude Code and want one memory across them.
- You want decisions with history and supersession, not just edited notes.
- You want handoffs and repeatable workflows without building them on the API.
Claude app memory fits when
- You want personal context in everyday chats with zero setup.
Managed Agents memory fits when
- You are building your own agent on the Claude Platform and want memory mounted as files with version history.
Sources
- Using Claude’s chat search and memory to build on previous context — Claude Help Center
- Claude’s memory works everywhere, and you decide what’s in it — Anthropic, August 25, 2026
- Claude release notes — Claude Help Center
- Memory for Claude Managed Agents — Anthropic, April 23, 2026
- Managed Agents memory — Claude Platform docs
- Dreams — Claude Platform docs
- New in Claude Managed Agents — Anthropic, May 19, 2026
- Memory tool — Claude Platform docs
- Claude Code memory — Claude Code docs
- Pricing — Anthropic
Frequently asked questions
Does ContxOS replace Claude’s memory?
No. Keep Claude’s memory for personal context. ContxOS adds a shared project record that also reaches Claude Code and other clients.
Is ContxOS made by Anthropic?
No. ContxOS is an independent product that connects to Claude through the Model Context Protocol.
Give your agent a memory that lasts
Connect ContxOS to Claude in about a minute. The free plan includes one project.
Product names belong to their owners. ContxOS is not affiliated with the companies named here.