MCP Memory Keeper
An MCP server that preserves Claude Code context between sessions: decisions, checkpoints and a file cache kept in local SQLite.
- 136 ★ on the repository
- Last commit
Extended access
Why this level
- The server reads and caches project file contents and keeps them in a local database.
- The git integration reads the project's branch history.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/mcp-memory-keeperDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads and caches project file contents and keeps them in a local database.
- The git integration reads the project's branch history.
This is third-party code. Review the repository files before installing.
What it does
Memory Keeper is an MCP server with long-term memory for Claude Code. The agent saves decisions, progress and findings with a category and priority, searches them, creates checkpoints and restores a session by key after the context is compacted. Data lives in a local SQLite database in `~/mcp-data/memory-keeper/`. Context items are grouped into channels derived from the git branch. Features include a file cache with change detection, time and regex filters, batch operations, links between items, and export and import. There are 38 tools, and the `minimal` (8), `standard` (22) and `full` (38) profiles are chosen with `TOOL_PROFILE` to keep context small.
Who it is for. Claude Code users in long sessions where context is compacted and decisions and progress get lost.
Good fit when
- Context nears its limit and a checkpoint with a restore key is created before compaction.
- Two Claude sessions run in parallel, one researching and one coding, and both read a shared memory.
- Architectural decisions and test results are saved with the `decision` or `progress` category.
- Project notes move to another machine through export and import.
Not a fit when
- You want memory that decides by itself what to save: here the user or a `CLAUDE.md` rule asks the agent to.
- The context contains secrets that must not be written to a local database.
- You need semantic search: search in this server is full-text.
Example request
Analyze the repository and save your analysis in memory-keeper under the auth-service channel, then create a checkpoint.MCP
- Transport
- stdio
- Authentication
- not required
Environment variables
DATA_DIR- Database directory, `~/mcp-data/memory-keeper/` by default.
TOOL_PROFILE- Tool profile: `minimal`, `standard`, `full` or a custom one from the config file.
MCP_MAX_TOKENS- Response size limit in tokens, 25000 by default.
Limitations
Memory fills at the user's request or by rules in `CLAUDE.md` and a custom command, and the server does not watch context by itself. Data is stored in `~/mcp-data/memory-keeper/`, and `DATA_DIR` changes the directory. All 38 tools are exposed by default and their descriptions take context space, while the `minimal` and `standard` profiles shrink the set. Response size is capped by `MCP_MAX_TOKENS` (25000 by default), and when exceeded only part of the items return. The `npx` option fetches the latest version every time, so pinning a version is manual.
How to disable. Run `claude mcp remove memory-keeper`. The database stays in `~/mcp-data/memory-keeper/` and is deleted by hand when the context is no longer needed.
FAQ
How do I restore context after a reset?
Ask Claude to create a checkpoint and report the key, then in a new session say restore from the key and paste it. Memory is shared, so another session works too.
Where is the data stored?
In `~/mcp-data/memory-keeper/`. The `DATA_DIR` variable sets another directory.
How do I reduce the number of tools?
Start the server with `TOOL_PROFILE=minimal` (8 tools) or `standard` (22). A custom profile is defined in `~/.mcp-memory-keeper/config.json`.
Does the server work in Claude Desktop?
Yes, a block with the `npx` command and the `mcp-memory-keeper` argument goes into the Claude Desktop configuration the same way as for Claude Code.
Do I need special rights to install it?
No. `claude mcp add memory-keeper npx mcp-memory-keeper` needs no manual build or compilation, and `npx` pulls native modules itself.
Discussion
No questions or reviews yet.
Related
MCP serversContextKeep gives agents persistent memory on SQLite with full-text search, categories and a web interface.
An MCP server that keeps project decisions in Redis or Valkey and brings them back in new sessions and after compaction.
Long-term memory for Claude Code, Cursor and Codex: the agent saves decisions and finds them again in later sessions
MemoMind MCP gives Claude Code project memory, with stored facts, events and conclusions visible in a local dashboard.
Local memory for agents: stores conversations and project files verbatim and retrieves them with semantic search over MCP
Local memory for an agent: hooks record facts from conversations and the MCP server brings them back in later sessions from one SQLite file
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MCP Memory Keeper (this card) | no | not required | 136 | |
| MCP Memory Service | no | not required | 2,090 | a self-hosted memory with semantic search and a knowledge graph; Memory Keeper is simpler, keeps data in local SQLite and searches by text |
| Claude-Mem | no | 113,534 | a plugin that saves session observations automatically; in Memory Keeper context is saved on the user's request or by a `CLAUDE.md` rule |
README in short
The README shows a quick start, an example workflow with `CLAUDE.md` and a custom command, the feature list, environment variables, tool profiles and a detailed description of every call.