KEPTA — the memory for AI systems
Local encrypted memory shared by Claude, Cursor and other MCP clients, in one file you own
- Pricing
- Free
- Platforms
- CLI, macOS, self-hosted
- Runs
- Locally
- Version
- 3.4.0
- Listing updated
- today
Similar MCP servers
About KEPTA — the memory for AI systems
KEPTA — the memory for AI systems is a free MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
K3PTA keeps what your AI learns in a single SQLCipher-encrypted database on your own machine, with the key held in the operating system keychain. Every connected client reads and writes the same memory, so a note saved from Claude can be recalled from Cursor or a script. There is no cloud and no account.
Retrieval combines full-text search, vectors and a knowledge graph, fused and reranked locally. The memory tracks what was true and when, so notes can be contradicted or superseded rather than just piled up, and a timeline command shows how facts changed. An import command brings in ChatGPT history, and a one-line setup command wires the server into clients such as Claude Desktop, Cursor, VS Code, Gemini CLI and Cline.
The core is source-available under BUSL-1.1 (converting to AGPL after four years), with a free desktop app for macOS and an HTTP API with 32 routes. Local writes stop at 5,000 active notes by default as a guard against runaway agents. Team and organisation features are sold separately as Enterprise. Needs Node 22.13 or newer.
Install
Run in a terminal
claude mcp add kepta -- npx -y kepta-mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"kepta": {
"command": "npx",
"args": [
"-y",
"kepta-mcp"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"kepta": {
"command": "npx",
"args": [
"-y",
"kepta-mcp"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"kepta": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"kepta-mcp"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"kepta": {
"command": "npx",
"args": [
"-y",
"kepta-mcp"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"kepta": {
"command": "npx",
"args": [
"-y",
"kepta-mcp"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
Tools this server provides (8)
Taken from the project documentation; the running server may differ.
- memory_search
- Search stored memories.
- memory_save
- Save a note.
- memory_update
- Update a note.
- memory_delete
- Delete a note.
- memory_list
- List notes.
- memory_graph
- Explore the knowledge graph.
- memory_consolidate
- Consolidate memories.
- memory_forget
- Forget a memory.
What it does
- Encrypted single file
- SQLCipher 4 with AES-256, key kept in the OS keychain.
- Hybrid retrieval
- Full-text, vector and graph search fused and reranked locally.
- One memory for every client
- Claude, Cursor, Gemini CLI, Cline and scripts share the same notes.
- Time-aware notes
- Track what was true and when, with superseded facts kept.
- Setup and import
- One command connects clients; ChatGPT exports can be imported.
- Desktop app
- A free window for browsing memory, graph and search provenance.
Good for
- Give every AI client the same persistent project memory
- Remember client details across chat sessions
- Move ChatGPT history into a private local store
- Script memory reads and writes through the HTTP API
Pricing
Free
Free core and desktop app; Enterprise by agreement
KEPTA — the memory for AI systems FAQ
Does it send data to the cloud?
No. The README says it is local, with no cloud and no account.
What does it cost?
The core and desktop app are free; KEPTA Enterprise for teams is licensed by agreement.
What Node version is needed?
Node 22.13 or newer.
Is there a note limit?
Locally the API stops accepting new notes at 5,000 active by default, adjustable with KEPTA_MAX_ACTIVE.
Is KEPTA — the memory for AI systems free?
Yes. KEPTA — the memory for AI systems is free to use.
How do I add KEPTA — the memory for AI systems to Claude Code?
Run this command in your terminal: claude mcp add kepta -- npx -y kepta-mcp
What are the best alternatives to KEPTA — the memory for AI systems?
The closest alternatives to KEPTA — the memory for AI systems are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and NeuroStack.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.



