memoriaBLOCK
Local-first, LLM-agnostic memory system — hybrid retrieval (vector + lexical + graph), multi-signal ranking. Runs on 4GB RAM, CPU-only, no cloud.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Local-first, LLM-agnostic memory infrastructure with parallel hybrid retrieval, multi-signal ranking, declarative type routing, temporal retrieval, and a plugin-based architecture.
Memoria provides a complete memory substrate without requiring a cloud service, API key, GPU, or even an LLM.
- CPU-only · ~4 GB RAM capable · No cloud dependency · No API keys required
- LLM optional · MIT licensed · Python
- CLI · TUI · GUI · API · MCP
- Pluggable retrieval and extension architecture
Memoria 1.0.0 is the first official public release.
Documentation
Full documentation, architecture notes, configuration reference, adapter documentation, retrieval details, and benchmark methodology:
https://kitzkatz.github.io/memoria/
What Is Memoria?
Memoria is memory infrastructure first — built to store, retrieve, rank, route, and manage information independently of whichever language model happens to consume it.
Rather than treating memory as a thin vector-search layer, Memoria combines multiple retrieval and ranking signals, with temporal retrieval running as a separately evaluable path that merges in at the end:
┌─────────────────────┐ │ Query │ └──────────┬──────────┘ │ ┌─────────────┴─────────────┐ │ │ Base Retrieval Temporal Retrieval │ │ ┌────────┼────────┐ │ │ │ │ │ FAISS BM25 Graph Temporal │ │ │ │ └────────┴────────┘ │ │ │ Fusion │ │ │ Base Ranking │
de38904514e1OBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add kitzkatz-memoria -- uvx kitzkatz-memoria
{
"mcpServers": {
"kitzkatz-memoria": {
"command": "uvx",
"args": [
"kitzkatz-memoria"
]
}
}
}Exposed tools (6)
4 read · 1 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
memory_delete | destructive | |
memory_fetch | read | |
memory_search | read | |
memory_store | read | |
memory_store_many | read | |
memory_update | write |
Trust audit
BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (5 observation(s))
- Shell
- declared (2 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (9)
data = pickle.load(f)
data = pickle.load(f)
benchmark_archive.tar.gz
memory_delete
.gitmodules
data = pickle.load(f)
data = pickle.load(f)
unique_id = hashlib.md5(f"{first}{last}{random.random()}".encode()).hexdigest()[:8]faiss-cpu, numpy, pydantic, fastapi, uvicorn, requests, sqlalchemy, rank-bm25
Gates applied: no_behavioural_pass.
de38904514e1full audit observations/trust-audit/mcp-server/kitzkatz__memoria.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | de38904514e1 | BLOCK | D | 69 | first audit |
Questions
What is the memoria MCP server?
Local-first, LLM-agnostic memory system — hybrid retrieval (vector + lexical + graph), multi-signal ranking. Runs on 4GB RAM, CPU-only, no cloud.
What tools does memoria expose?
6 in total: 4 read-only, 1 that write, and 1 that can delete or overwrite (memory_delete). Every one is listed on this page with its risk.
Is memoria safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (69/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does memoria need?
No credential environment variables were found in its source, so it appears to need none.
How current is this page?
The grade is for one exact copy of the source (de38904514e1), read on 2026-10-08. The repository is watched and re-audited when it changes.