The Memory MCP server provides persistent memory and full-text session search for AI coding assistants, enabling them to remember context across sessions and search historical conversations.
Memory Management
save_memory – Persist notes, decisions, patterns, or preferences with optional tags and context.
search_memory – Full-text keyword search across saved memories, ranked by relevance, with optional tag filtering.
list_memories – Browse recent saved memories, optionally filtered by tag.
delete_memory – Remove a specific saved memory by its ID.
Session Search & Retrieval
list_sessions – Browse past AI coding sessions, filterable by source (e.g., claude_code, omp) or project path, with pagination.
get_session – Retrieve the full conversation from a specific session, including messages, assistant responses, and tool usage.
search_sessions – Full-text keyword search across all historical session messages, thinking blocks, and tool usage, ranked by relevance.
refresh_sessions – Re-scan session directories to index new or changed files, keeping session history up to date.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Memory MCP what did we decide about the database schema last week?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Persistent memory and full-text session search for AI coding assistants, exposed as an MCP server.
AI coding assistants forget everything between sessions. Architecture decisions, user preferences, project context, what you debugged last Tuesday -- gone. You re-explain the same things constantly.
Memory MCP fixes this with two capabilities:
Explicit memory -- save notes, decisions, patterns, and preferences that persist across sessions. Your assistant remembers what you told it.
Session search -- full-text search across your entire conversation history. Find that thing you discussed three weeks ago without scrolling through logs.
No database servers. No background processes. No cloud. One SQLite file on your machine.
Related MCP server: MCP Vector Memory
Source | Location | Format |
| JSONL (streamed content blocks) | |
Claude Code history |
| JSONL (survives session file pruning) |
| SQLite (sessions, messages, parts tables) | |
| JSONL (event-per-line) | |
| JSONL (rollout events) | |
| JSON (chat sessions) | |
| JSON (conversations) | |
LM Studio API logs |
| JSONL (via |
Adding a new source requires one parser file and a registry entry. See Adding a new source.
Requires Python 3.11+ with SQLite FTS5 support (included in standard Python builds).
pip install -e .Or run directly with uv (no install needed):
uv run --directory /path/to/memory_mcp python -m memory_mcpAdd to your MCP client config (e.g., ~/.claude/mcp.json or project-level .mcp.json):
With pip install:
{
"mcpServers": {
"memory": {
"command": "memory-mcp"
}
}
}With uv (no install):
{
"mcpServers": {
"memory": {
"command": "uv",
"args": ["run", "--directory", "/path/to/memory_mcp", "python", "-m", "memory_mcp"]
}
}
}Tool | Description |
| Persist a note with optional tags and context. Survives across all future sessions. |
| Full-text search across saved memories. Keyword-based, ranked by relevance. |
| Browse recent memories, optionally filtered by tag. |
| Remove a memory by ID. |
Tool | Description |
| Browse past sessions. Filter by source ( |
| Retrieve the full conversation from a specific session. |
| Get tool calls and results for a session, optionally filtered by tool name. |
| Full-text search across all session messages, thinking blocks, and tool usage. |
| Re-scan session directories and index new or changed files. |
| Manually trigger a full sync push + pull cycle. |
Memory MCP can sync sessions and memories across multiple machines via a self-hosted sync server. When configured, each machine pushes its local data to a central PostgreSQL database and pulls data from other machines.
Deploy the sync server with PostgreSQL + systemd. See
DEPLOY.md for the Proxmox/no-Docker playbook.
Current lab shape:
memory-mcp app VM: FastAPI service on :8000
pg2026 DB VM: PostgreSQL 18 + pgvector
Create an API key:
python3 -c "import secrets; print(secrets.token_hex(32))"Insert the SHA-256 hash into PostgreSQL:
INSERT INTO users (id, name, api_key_hash, created_at)
VALUES (gen_random_uuid(), 'austin', '<sha256-of-api-key>', now());Configure each machine with environment variables:
export MEMORY_MCP_SYNC_URL=http://your-server:8000
export MEMORY_MCP_SYNC_KEY=your-secret-keyRestart memory-mcp — the sync engine starts automatically.
On the first configured sync, existing rows in ~/.memory_mcp/memory.db
are assigned this machine's UUID and uploaded; no separate SQLite export is
needed.
Offline-first: all reads go to local SQLite. Sync is a background process — your tools are never blocked waiting for the network.
Push: pending sessions and memories are POSTed to the server after
each scan cycle and after each save_memory call.
Pull: the server returns items authored by other machines since the
last pull. Sessions use INSERT OR IGNORE (idempotent); memories use
last-write-wins conflict resolution by updated_at.
Machine identity: each host generates a persistent UUID on first run
(~/.memory_mcp/machine_id). This UUID is the sync key.
No env vars = local-only: if MEMORY_MCP_SYNC_URL and
MEMORY_MCP_SYNC_KEY aren't set, the sync engine never starts and
behavior is identical to v0.3.0.
Tool | Description |
| Manually trigger a full push + pull cycle. Returns a summary. |
On startup, Memory MCP yields its tool list to the MCP client immediately (<500 ms cold) and runs the initial session scan in a background task. The embedding model loads lazily on the first semantic search call — keyword search and saved memories work without it. Subsequent startups skip files whose mtime hasn't changed.
Database location: ~/.memory_mcp/memory.db (override with MEMORY_MCP_DB env var)
Session sources: auto-detected from standard locations (extend with MEMORY_MCP_SOURCES env var, format: type:path;type:path)
Indexing: incremental by file mtime, parallelized across 8 threads
Search: FTS5 with BM25 ranking, prefix matching, phrase support; optional vector search via sqlite-vec + fastembed (BAAI/bge-small-en-v1.5) when semantic=true is passed
Startup: non-blocking — heavy work (scan, model load, vector backfill) runs after the server is already responding to tool calls
Create memory_mcp/parsers/your_source.py implementing the SessionParser protocol:
source_type: str attribute
parse_file(path: str) -> ParsedSession | None method
Register it in memory_mcp/parsers/__init__.py
Add directory detection in memory_mcp/config.py
See parsers/claude_code.py or parsers/omp.py for examples.
python tests/test_e2e.py # end-to-end: spawns server, exercises all 10 tools
python tests/test_startup.py # startup contract: cold Popen -> tools/list under 1.5stest_e2e.py starts the MCP server as a subprocess, exercises all 10 tools over the stdio protocol, and asserts tool responses. test_startup.py enforces the v0.3.0 startup contract — if an eager import or pre-yield blocking call regresses startup speed, it fails immediately. Both use throwaway databases so your real data is untouched.
memory_mcp/
server.py # FastMCP entry point, lifespan yields fast then runs scan + sync in background
readiness.py # Lazy embedder + scan/backfill coordination
config.py # Auto-detects session dirs, DB path, sync settings
db.py # SQLite + FTS5 + sqlite-vec schema, all queries, sync triggers
embeddings.py # Lazy fastembed wrapper (BAAI/bge-small-en-v1.5)
scanner.py # Walks session dirs, dispatches to parsers, parallel indexing
machine_id.py # Persistent machine UUID for cross-machine sync
client.py # HTTP client for sync server (stdlib urllib, zero-dependency)
sync_engine.py # Background push/pull sync loop
parsers/
base.py # ParsedSession / ParsedMessage dataclasses, SessionParser protocol
claude_code.py # Claude Code JSONL parser (merges streamed assistant blocks)
claude_history.py # Claude Code history.jsonl parser (one file, many sessions)
omp.py # OMP JSONL parser
opencode.py # OpenCode SQLite parser (reads DB directly, read-only)
tools/
memory.py # save_memory, search_memory, list_memories, delete_memory
sessions.py # list_sessions, get_session, search_sessions, refresh_sessions
hosted/
server.py # FastAPI sync server (REST API)
models.py # SQLAlchemy models (PostgreSQL + pgvector)
auth.py # Bearer API key authenticationMIT
Delete a specific memory by its ID.
| Name | Required | Description | Default |
|---|---|---|---|
| memory_id | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states the tool deletes a memory, implying a destructive mutation, but doesn't cover critical aspects like whether deletion is permanent, requires specific permissions, has side effects (e.g., on related sessions), or returns confirmation data. For a destructive tool, this leaves significant gaps in understanding its behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, direct sentence with zero wasted words. It front-loads the key action ('Delete') and resource ('memory'), making it immediately scannable. Every word earns its place by contributing essential information without redundancy or fluff.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (destructive operation with 1 parameter) and lack of annotations, the description is minimally adequate. The presence of an output schema reduces the need to explain return values, but the description doesn't address behavioral risks or usage context. It covers the basic 'what' but misses the 'how' and 'when,' leaving room for improvement in safety and clarity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0% description coverage, so the description must compensate. It adds meaning by specifying that 'memory_id' identifies 'a specific memory,' clarifying the parameter's role. However, it doesn't explain what a memory ID is (e.g., format, source) or constraints (e.g., valid ranges), leaving the schema's bare type ('integer') as the only guidance. This partial compensation earns a baseline score.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Delete') and target resource ('a specific memory by its ID'), making the purpose immediately understandable. It distinguishes from siblings like 'list_memories' and 'save_memory' by focusing on deletion rather than retrieval or creation. However, it doesn't specify what constitutes a 'memory' in this context, which slightly limits specificity.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention prerequisites (e.g., needing an existing memory ID), exclusions, or comparisons to siblings like 'save_memory' for creation or 'list_memories' for retrieval. Without this context, users must infer usage from the tool name alone.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Retrieve the conversation from a specific session. Shows the full message flow including user messages, assistant responses, and tool usage.
| Name | Required | Description | Default |
|---|---|---|---|
| session_id | Yes | ||
| limit | No | ||
| offset | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It describes what the tool returns ('full message flow'), but doesn't disclose behavioral traits such as authentication needs, rate limits, error handling, or whether it's a read-only operation (though 'retrieve' implies reading). For a tool with no annotations, this leaves significant gaps in understanding its behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences that are front-loaded and efficient. The first sentence states the core purpose, and the second adds valuable detail about the content retrieved. There is no wasted language, and every sentence earns its place by enhancing clarity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has an output schema (which likely describes the return structure), the description doesn't need to explain return values. However, with no annotations, 3 parameters (one required), and 0% schema description coverage, the description is incomplete—it lacks parameter details and behavioral context. It's minimally adequate but has clear gaps for a retrieval tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the schema provides no descriptions for parameters. The description adds no information about parameters beyond what the schema names imply (e.g., session_id, limit, offset). It doesn't explain what a session_id is, how limit/offset affect pagination, or any constraints. With low coverage, the description fails to compensate, resulting in a baseline score.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'retrieve' and the resource 'conversation from a specific session', and specifies the content includes 'user messages, assistant responses, and tool usage'. It distinguishes from siblings like list_sessions or search_sessions by focusing on retrieving a specific session's content rather than listing or searching. However, it doesn't explicitly contrast with siblings like refresh_sessions, so it's not a perfect 5.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention when to use get_session over list_sessions or search_sessions, nor does it specify prerequisites or exclusions. The context is implied by the action of retrieving a specific session, but no explicit usage rules are given.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List recent saved memories, optionally filtered by tag.
| Name | Required | Description | Default |
|---|---|---|---|
| tag | No | ||
| limit | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It mentions 'recent saved memories' but doesn't specify what 'recent' means (e.g., time frame, ordering), whether results are paginated, or any rate limits or permissions required. The description adds minimal context beyond the basic operation, leaving key behavioral traits undefined.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that front-loads the core purpose ('List recent saved memories') and adds an optional feature ('optionally filtered by tag') without unnecessary details. Every word earns its place, making it highly concise and well-structured for quick understanding.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's low complexity (2 optional parameters) and the presence of an output schema (which handles return values), the description is minimally adequate. However, with no annotations and 0% schema coverage, it lacks details on behavioral aspects like ordering, pagination, or error handling. It meets the baseline for a simple list tool but leaves gaps in usage and parameter context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the description must compensate for undocumented parameters. It mentions optional filtering by 'tag' but doesn't explain tag semantics or format. It doesn't address the 'limit' parameter at all, leaving its purpose and constraints unclear. The description adds some meaning for 'tag' but fails to cover 'limit', resulting in incomplete parameter guidance.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('list') and resource ('recent saved memories') with an optional filtering capability ('by tag'). It distinguishes from siblings like 'search_memory' by focusing on listing rather than searching, though it doesn't explicitly contrast with 'get_session' or 'list_sessions'. The purpose is specific and actionable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for retrieving memories with optional tag filtering, but provides no explicit guidance on when to use this tool versus alternatives like 'search_memory' or 'list_sessions'. It mentions the filtering capability but doesn't clarify scenarios where listing is preferred over searching or other sibling tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List recent AI coding sessions. Filter by source ('claude_code' or 'omp') or by project path substring.
| Name | Required | Description | Default |
|---|---|---|---|
| source | No | ||
| project | No | ||
| limit | No | ||
| offset | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It mentions filtering capabilities but lacks details on permissions, rate limits, pagination behavior (beyond implied by limit/offset), or what 'recent' means. For a list tool with no annotation coverage, this leaves significant gaps in understanding how the tool behaves operationally.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that front-loads the core purpose and immediately details filtering options. Every word contributes directly to understanding the tool's functionality, with no wasted text or redundancy. It's appropriately sized for the tool's complexity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given 4 parameters with 0% schema coverage and no annotations, the description partially compensates by explaining two parameters. An output schema exists, so return values need not be described. However, for a list tool with filtering and pagination, the description lacks completeness in behavioral aspects like default sorting, error handling, or full parameter coverage, making it adequate but with clear gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the schema provides no parameter descriptions. The description adds value by explaining that 'source' accepts 'claude_code' or 'omp' and 'project' filters by path substring, covering 2 of 4 parameters. However, it doesn't address 'limit' or 'offset', leaving half the parameters without semantic context. This partial compensation results in a baseline score.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('List') and resource ('recent AI coding sessions'), making the purpose specific and understandable. It distinguishes from siblings like 'get_session' (singular) and 'search_sessions' (broader search), though it doesn't explicitly name alternatives. The purpose is not vague or tautological.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for filtering by source or project path, suggesting when to use it for basic filtering. However, it doesn't provide explicit guidance on when to choose this tool over alternatives like 'search_sessions' or 'refresh_sessions', nor does it mention any exclusions or prerequisites. The guidance is limited to implied context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Scan for new or updated session files and index them. Run this if recent sessions aren't showing up in search results.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It discloses the tool's purpose (scanning and indexing) and effect (making sessions searchable), but lacks details on permissions needed, rate limits, whether it's idempotent, or what 'indexing' entails technically. It adds some context but is incomplete for behavioral transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, front-loaded with the core action and followed by usage guidance. Every word earns its place—no redundancy or fluff. Efficiently conveys purpose and when to use in minimal text.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given 0 parameters, no annotations, and an output schema (which handles return values), the description is reasonably complete. It explains what the tool does and when to use it, but could improve by adding more behavioral context (e.g., side effects, time to index). The output schema reduces the need for return value details.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are 0 parameters, and schema description coverage is 100% (empty schema). The description doesn't need to explain parameters, so it meets the baseline of 4 for tools with no parameters, as it can't add value beyond what the schema already provides (none).
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('Scan for new or updated session files and index them') and resource ('session files'), distinguishing it from siblings like get_session (retrieve one), list_sessions (list existing), and search_sessions (search indexed sessions). It specifies the action of indexing, which siblings don't cover.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly states when to use: 'Run this if recent sessions aren't showing up in search results.' This provides a clear trigger condition and implies an alternative (waiting for automatic indexing or using search_sessions without refresh). It differentiates from siblings by addressing indexing lag.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Save a note to persistent memory. Use this to remember important context, decisions, patterns, or user preferences across sessions.
| Name | Required | Description | Default |
|---|---|---|---|
| content | Yes | ||
| tags | No | ||
| context | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It discloses that the tool saves data persistently across sessions, which is a key behavioral trait. However, it lacks details on permissions, rate limits, error handling, or how the saved data is structured or retrieved, leaving gaps in behavioral understanding.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately sized and front-loaded, consisting of two concise sentences. The first sentence states the core purpose, and the second provides usage context, with no wasted words or redundant information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (a write operation with 3 parameters), lack of annotations, and presence of an output schema, the description is moderately complete. It explains the tool's purpose and usage context but misses details on parameters, behavioral constraints, and how it interacts with sibling tools, making it adequate but with clear gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema description coverage is 0%, so the description must compensate. It does not mention any parameters explicitly, failing to add meaning beyond the input schema. The baseline is 3 because the schema covers all parameters (content, tags, context) with titles and types, but the description provides no additional semantic context for their use.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with a specific verb ('save') and resource ('note to persistent memory'), distinguishing it from sibling tools like delete_memory, list_memories, and search_memory. It specifies that this is for writing data that persists across sessions, unlike get_session or list_sessions which are read-only.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear context on when to use this tool ('to remember important context, decisions, patterns, or user preferences across sessions'), but does not explicitly state when not to use it or name alternatives. For example, it doesn't clarify whether to use save_memory versus other tools for similar purposes, though the context implies it's for persistent storage.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search saved memories. Use short keywords, not natural language phrases — each word is matched independently and ranked by relevance.
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | ||
| tags | No | ||
| limit | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It adds useful context about search behavior ('each word is matched independently and ranked by relevance'), which goes beyond the basic purpose. However, it doesn't cover other important aspects like authentication needs, rate limits, error conditions, or what the output contains, leaving gaps for a tool with no annotation support.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is perfectly concise and well-structured in a single sentence that front-loads the purpose and follows with specific usage guidance. Every word serves a purpose with no wasted text, making it easy to parse quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given that there's an output schema (which handles return values) but no annotations and 0% schema description coverage, the description provides adequate basic purpose and usage guidance but falls short on parameter explanations and behavioral details. For a search tool with 3 parameters, it should ideally provide more complete parameter semantics to compensate for the schema gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema description coverage is 0%, meaning none of the parameters have descriptions in the schema. The tool description mentions 'short keywords' which relates to the 'query' parameter, but doesn't explain the semantics of 'tags' or 'limit' parameters. This partial coverage doesn't adequately compensate for the complete lack of schema descriptions for a 3-parameter tool.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with the verb 'Search' and resource 'saved memories', making it immediately understandable. It distinguishes itself from siblings like 'list_memories' by specifying a search functionality, though it doesn't explicitly contrast with 'search_sessions' which might have similar search mechanics but different targets.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear context on how to use the tool effectively ('Use short keywords, not natural language phrases'), which helps guide proper invocation. However, it lacks explicit guidance on when to choose this tool over alternatives like 'list_memories' or 'search_sessions', leaving some ambiguity in sibling differentiation.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search across all session messages. Use short keywords, not natural language phrases — each word is matched independently and ranked by relevance.
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | ||
| limit | No | ||
| offset | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It adds useful context about search behavior ('each word is matched independently and ranked by relevance'), but doesn't cover aspects like permissions, rate limits, or response format. This is adequate but has clear gaps for a search tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is highly concise and front-loaded: it starts with the core purpose, followed by specific usage advice. Every sentence adds value without redundancy, making it efficient and well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (search with pagination), no annotations, and an output schema (which handles return values), the description is partially complete. It covers search behavior but misses key details like pagination semantics and lacks sibling differentiation, leaving room for improvement.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the description must compensate for undocumented parameters. It only mentions the 'query' parameter implicitly ('Use short keywords'), but doesn't explain 'limit' or 'offset' (pagination), nor does it clarify the query syntax or ranking details. This leaves significant gaps in parameter understanding.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Search across all session messages.' It specifies the resource (session messages) and verb (search), but doesn't explicitly differentiate from sibling tools like 'search_memory' or 'list_sessions,' which would require a 5.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides some guidance on how to use the tool effectively ('Use short keywords, not natural language phrases'), but doesn't specify when to choose this tool over alternatives like 'search_memory' or 'list_sessions.' It implies usage context without explicit exclusions or comparisons.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v0.1.0delete_memoryget_sessionlist_memorieslist_sessionsrefresh_sessionssave_memorysearch_memorysearch_sessionsEach tool has a clearly distinct purpose with no overlap: delete_memory removes a memory, get_session retrieves a session's conversation, list_memories lists memories, list_sessions lists sessions, refresh_sessions indexes new sessions, save_memory stores a memory, search_memory searches memories, and search_sessions searches sessions. The descriptions reinforce these distinct roles, making misselection unlikely.
All tools follow a consistent verb_noun pattern using snake_case: delete_memory, get_session, list_memories, list_sessions, refresh_sessions, save_memory, search_memory, and search_sessions. This predictable naming scheme enhances readability and usability across the toolset.
With 8 tools, the server is well-scoped for managing memories and sessions. Each tool serves a clear purpose, such as CRUD operations for memories (save, list, search, delete) and sessions (list, get, search, refresh), avoiding bloat while covering essential functionality for the domain.
The toolset provides strong coverage for memory and session management, including create (save_memory), read (list_memories, get_session), delete (delete_memory), and search (search_memory, search_sessions). A minor gap exists in updating memories or sessions, but agents can work around this by deleting and re-saving, and the refresh_sessions tool handles session updates indirectly.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Persistent memory for AI agents. Search, store, and recall across sessions.
Persistent memory for AI agents. Search and store durable facts, preferences and decisions.
Shared memory for coding agents. Stop re-explaining your codebase every session.
Persistent memory for AI agents — verbatim conversations, searchable by meaning.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/nerdyaustin/memory_mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
