memalyze
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MEMALYZE_DB | No | SQLite database path | ./memalyze.db |
| MEMALYZE_ACTOR | No | Actor recorded in audit | local |
| MEMALYZE_AUDIT | No | Whether to append audit rows | on |
| MEMALYZE_TENANT | No | Default tenant for tool calls | default |
| MEMALYZE_HTTP_ADDR | No | HTTP listen address, e.g. :8080 for Streamable HTTP; empty means stdio | |
| MEMALYZE_HTTP_TOKEN | No | Bearer token for HTTP; required when binding off-loopback | |
| MEMALYZE_KEEP_ALIVE | No | Ollama model residency | 30m |
| MEMALYZE_OLLAMA_URL | No | Local embedding endpoint | http://127.0.0.1:11434 |
| MEMALYZE_EMBED_MODEL | No | Embedding model | qwen3-embedding:0.6b |
| MEMALYZE_REDACT_MODE | No | Redaction mode: enforce, dry-run, or off | enforce |
| MEMALYZE_REDACT_RULES | No | Path to a JSON rule pack; defaults to built-in rules | (built-in) |
| MEMALYZE_QUERY_TIMEOUT | No | Vector-path budget before degrading | 3s |
| MEMALYZE_AUDIT_QUERY_PLAIN | No | Store query text in audit; by default only SHA-256 and length are stored | off |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| echoA | Echo the given text back unchanged. Used to verify the MCP wiring. |
| memory_addA | Store a memory for later retrieval. Returns duplicate=true when identical content already exists (no new record), and embedding_failed=true when the memory was saved but vectorization failed (keyword search still works). |
| memory_deleteA | Soft-delete a memory by id. Already-deleted ids are idempotent; an unknown id returns an error instead of pretending to succeed. |
| memory_searchA | Search memories with vector search first and keyword fallback. degraded=true means the vector path was unavailable; degrade_reason explains why. An empty query lists recent memories. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: echo is a wiring check, while memory_add, memory_delete, and memory_search map to separate lifecycle operations on a single resource. Descriptions reinforce the boundaries (e.g., soft-delete vs. store vs. hybrid search).
The three core tools follow a clean memory_verb snake_case pattern. echo is the only deviation, but as a conventional diagnostic utility it does not create confusion.
Four tools is slim but reasonable for a focused memory store; each earns its place except arguably echo, which is a wiring diagnostic rather than a memory operation. Not thin enough to be a mismatch.
Add, search (with list-recent fallback), and delete cover the core lifecycle, but there is no get-by-id or update/edit operation, so an agent cannot correct or inspect a specific memory without going through search. These are minor workaround-able gaps assuming memories are effectively immutable.