Memors
A local SQLite knowledge base for agents with hybrid search and provenance on every result
- Pricing
- Open source
- Platforms
- macOS, Linux, CLI
- Runs
- Locally
- Version
- 1.6.10
- Listing updated
- today
Similar MCP servers
About Memors
Memors is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
Memors-mcp is a single Go binary that keeps one SQLite file per knowledge base on your machine. Documents and facts go in with provenance, and search returns one ranked list fused from keyword (FTS5 with BM25), exact-identifier and local embedding arms, with structural entity and graph arms added when a question names related things. With response_format set to explain, each result shows per-arm ranks, recency factor and why the list was cut.
Facts carry evidence and validity dates, corrections supersede old facts instead of overwriting them, and as_of queries answer with what was believed on a date. Tool calls can retire only records written by tools, and promoting trust needs a human. Curated pages can be submitted from compact proposals and are marked stale when their sources change.
It began as a Go rewrite of obra/private-journal-mcp. The only outbound network call is the one-time embedding model download from Hugging Face. A read-only web UI on loopback, terminal commands and a macOS menu-bar app are included for people.
Install
Run in a terminal
claude mcp add memors -- npx -y https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gzAdd to claude_desktop_config.json
{
"mcpServers": {
"memors": {
"command": "npx",
"args": [
"-y",
"https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"memors": {
"command": "npx",
"args": [
"-y",
"https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"memors": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"memors": {
"command": "npx",
"args": [
"-y",
"https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"memors": {
"command": "npx",
"args": [
"-y",
"https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
MEMORS_KBName of the knowledge base to open; one SQLite file per name under $MEMORS_HOME/kb/.MEMORS_HOMEBase directory for knowledge bases and settings (default ~/.memors-mcp).MEMORS_MODELEmbedding model id from `memors-mcp model ls` (default granite-small-r2).
Tools this server provides (10)
Taken from the project documentation; the running server may differ.
- ingest
- Stores a document as markdown, creating a new revision when content changes.
- search
- Finds passages by words, identifiers and meaning in one fused list.
- read
- Dereferences a memo:// address.
- remember
- Records one atomic fact with evidence and validity dates.
- forget
- Retires a document or fact with a reason.
- promote
- Asks to raise a record's trust.
- explore
- Walks the graph index from one named thing.
- compact
- Proposes tidying work such as stale pages and duplicate names.
- submit
- Hands back a page, conflict decision or merge decision.
- status
- Shows namespaces, model, pending vectors and jobs.
What it does
- Hybrid ranked search
- Keyword, exact-identifier, semantic and graph arms fused into one list.
- Explainable results
- Per-arm ranks, recency factor and a query trace on request.
- Versioned facts
- Record facts with evidence and validity dates; supersede rather than overwrite.
- memo:// addresses
- Dereference passages, sections and documents under a token budget.
- Human-gated trust
- Tool calls cannot raise trust; a person promotes records.
- Local-only storage
- One SQLite file per knowledge base with no cloud dependency.
Good for
- Give a coding agent a durable, searchable project memory
- Track facts that change over time and ask what was believed on a date
- Find passages by exact identifier like ERR_CONN_RESET
- Review why a search result ranked where it did
Pricing
Open source
Memors FAQ
Does anything leave my machine?
Only the first-start embedding model download from Hugging Face.
Where is data stored?
In a SQLite file per knowledge base under ~/.memors-mcp/kb/ by default.
Which embedding model is default?
granite-small-r2, run locally, with alternatives listed by memors-mcp model ls.
Can a human browse it?
Yes, through memors-mcp ui, a read-only web page on loopback.
Is Memors free?
Yes. Memors is open source, so you can use and self-host it for free.
How do I add Memors to Claude Code?
Run this command in your terminal: claude mcp add memors -- npx -y https://github.com/kKEo/memors/releases/download/v1.6.10/memors-mcp_1.6.10_linux_amd64.tar.gz
What are the best alternatives to Memors?
The closest alternatives to Memors 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.



