Claude Memory Connector: Add Persistent Work Context

A memory connector for Claude is an external tool that feeds a structured work record into Claude at the start of a session—filling the gap that built-in conversation memory leaves when you start a new chat, switch projects, or move to a different AI app. The connector is typically an MCP server (a standard way for an AI tool to call an outside service) that returns the current rules, where the work was left, and what is in progress.

This is different from Claude's own memory feature. Built-in memory records personal preferences and facts about you; an external connector supplies a live record of the work. Tools like fenbs act as that connector, keeping projects, tasks, Decisions and rules, and Where we left off outside any one AI app so the context survives session changes.

What Claude's built-in memory does

Claude.ai has a memory feature that stores facts across conversations: your preferences, recurring context about your work, and information you have asked it to remember. This is personal memory—it stays in your Claude account and applies to conversations you start there.

Built-in memory is useful for personal context that applies across many topics, such as "always write in concise paragraphs" or "I work in the Eastern time zone." It is less suited to project-level context that changes often: the current state of a task, a decision made three days ago, or a rule that applies to one project only.

Built-in memory is also not shared. A colleague's Claude session does not see your stored memories. If you switch to Claude Code, Cursor, or ChatGPT, that memory does not travel with you. For the limits and behavior of Claude's built-in memory, see the Claude memory management guide.

What Claude Projects contribute

Claude Projects let you add standing instructions and reference files to a named project in Claude.ai. Those instructions load at the start of every conversation in that project without you having to paste them in.

Projects are better suited to slow-changing context—an overview of the project, a style guide, a list of conventions—than to fast-changing work status. They also stay in Claude.ai; they are not visible when you switch to Claude Code or a different AI app. See Claude project memory for a full breakdown of what lives there.

What an external memory connector adds

An external memory connector sits outside Claude and serves structured, updatable context to any AI app that connects to it. Instead of copying and pasting status updates into every session, you maintain one record in the connector and point Claude at it.

The difference from built-in memory and Projects is that the connector is the source of truth: Claude reads it, adds to it, and saves progress back to it. The record is maintained by people who review it, correct it, and decide what it says. An assistant cannot decide or sign off on anything stored there—it can record what a person decided, but the person makes the choice.

• Rules and decisions are written by people and read by every session that connects, regardless of which AI app is being used.
• Where we left off is saved by the last assistant to work on a task, so the next session—in any connected app—starts from an accurate picture.
• Task status, open questions, and context notes all update in the same record rather than accumulating in separate chat histories.
• Teammates can read and add to the same record, making it a shared work memory, not just a personal one.

Why an external connector works across models

Built-in Claude memory stays in Claude. A Claude Projects instruction file stays in Claude.ai. An external connector connected via MCP is model-agnostic: Claude, ChatGPT in developer mode, Cursor, Codex CLI and Gemini CLI can all call the same server if it supports MCP.

This matters when you switch tools. If you start a session in Claude Code and finish it in Cursor, the work context in an external connector survives that switch. The Rules and Where we left off are in the connector, not in Claude's memory. The next assistant reads them fresh. See Claude Code memory for the specifics of how Claude Code handles its own context alongside external connectors.

Choosing which context to keep where

Use Claude's built-in memory for personal, persistent preferences that apply across everything you do: communication style, time zone, background context. Use Claude Projects for reference documents and standing instructions specific to one Claude.ai project. Use an external connector for the live work record: task status, decisions, rules, and progress that should be available to any assistant, any teammate, and any session.

The three layers complement each other. Built-in memory knows who you are. Projects know the standing context of one body of work. An external connector knows where the work stands right now. Mixing all three into the same place creates a record that is hard to maintain and harder to trust.

How to do this with fenbs

1. Create a free account and add your first project. Give it a short description and add the current tasks.
2. Write the standing rules for that project in Decisions and rules—instructions every assistant should follow, with the person who decided them.
3. Open Integrations and follow the setup guide for your AI app. fenbs connects over MCP; the guide covers Claude Code, Claude.ai and other supported apps.
4. At the start of each session, ask Claude to call the context tool. It returns the project context in the right order: Rules, Where we left off, In progress, then any notes.
5. Before you close the session, ask Claude to save progress. It records what was done, what remains and the next step, so the next session—whether in Claude or another app—starts accurately.