AI Memory MCP
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., "@AI Memory MCPsearch my memories for the project plan"
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.
AI Memory MCP
AI Memory MCP gives agents one stable interface for durable memory. The server combines exact, lexical, semantic, and graph search results.
Markdown files are the source of truth. The system derives SQLite indexes and Graphify graphs from Markdown.
Architecture decision
The system is built around Graphify. Graphify remains a replaceable provider behind the stable AI Memory MCP interface.
The MCP facade owns scope, ranking, evidence, health, feedback, and refresh control. Clients do not depend on Graphify commands, files, or response formats.
This boundary permits a future graph-provider change without a client configuration change. Read the complete architecture guide for the design rules.
Related MCP server: BuildAutomata Memory MCP Server
System architecture
flowchart LR
subgraph Clients["MCP clients"]
Codex[Codex]
Claude[Claude]
Copilot[Copilot]
OpenCode[OpenCode]
VSCode[VS Code]
end
subgraph Repository["AI Memory MCP repository"]
MCP["MCP facade<br/>Eight public tools"]
Service["MemoryService<br/>Policy and orchestration"]
Retrieval["RetrievalEngine<br/>Scope, fusion, and reranking"]
Indexer["Memory indexer<br/>Validation and snapshots"]
Adapter["GraphifyAdapter<br/>Replaceable provider boundary"]
Skill["Canonical ai-memory skill"]
end
subgraph Authority["Canonical authority"]
Markdown["Markdown memory<br/>Durable source of truth"]
end
subgraph Derived["Derived state"]
SQLite["Versioned SQLite<br/>FTS5 and semantic vectors"]
Graph["Graphify graph<br/>Nodes, edges, and paths"]
end
Clients --> MCP
MCP --> Service
Service --> Retrieval
Retrieval --> SQLite
Retrieval --> Adapter
Adapter --> Graph
Markdown --> Indexer
Indexer --> SQLite
Markdown --> Graph
Skill -. guides .-> ClientsMain components
Component | Responsibility |
Markdown memory | Stores canonical durable records. |
Memory indexer | Validates records and publishes versioned SQLite snapshots. |
SQLite FTS5 | Supplies exact and lexical candidates. |
Local semantic index | Supplies paraphrase candidates without an external API. |
Graphify adapter | Supplies relationships, neighbors, and paths. |
Retrieval engine | Applies scope, RRF fusion, reranking, and context expansion. |
MCP facade | Supplies the stable public tools and evidence packets. |
Canonical skill | Gives agents the memory workflow and safety rules. |
Client installer | Registers the MCP server and repository-linked skill stubs. |
Query architecture
memory_search is the normal recall tool.
The retrieval engine plans and combines all provider work internally.
sequenceDiagram
participant Agent
participant MCP as MCP facade
participant Service as MemoryService
participant Engine as RetrievalEngine
participant SQLite as SQLite index
participant Graphify as GraphifyAdapter
Agent->>MCP: memory_search query and scope
MCP->>Service: Validate tool input
Service->>Engine: Search request
Engine->>Engine: Resolve scope and ticket identifiers
Engine->>SQLite: Get scoped lexical candidates
SQLite-->>Engine: Ranked lexical hits
Engine->>SQLite: Get scoped semantic vectors
SQLite-->>Engine: Ranked semantic hits
Engine->>Graphify: Rank scoped candidate paths
Graphify-->>Engine: Relationship signals
Engine->>Engine: Fuse results with RRF
Engine->>Engine: Deduplicate and rerank
Engine->>SQLite: Expand bounded context
Engine-->>Service: Evidence packet with source paths
Service-->>MCP: Structured result
MCP-->>Agent: Cited memory evidenceThe engine applies scope before it ranks each provider result. Exact identifiers receive bounded bonuses during reranking.
Graph traversal contributes one retrieval signal. It does not replace lexical or semantic retrieval.
Refresh architecture
An index refresh updates SQLite after a normal Markdown change. A full refresh also rebuilds the Graphify graph.
flowchart TD
Change[Canonical Markdown change] --> Mode{Refresh mode}
Mode -->|index| IndexStage[Build staged SQLite snapshot]
Mode -->|full| GraphStage[Build staged Graphify data]
GraphStage --> GraphValidate{Graph validation}
GraphValidate -->|pass| GraphPublish[Publish Graphify graph]
GraphValidate -->|fail| GraphKeep[Keep the last satisfactory graph]
GraphPublish --> IndexStage
IndexStage --> IndexValidate{Index validation}
IndexValidate -->|pass| Pointer[Publish the current-index pointer]
IndexValidate -->|fail| IndexKeep[Keep the last satisfactory index]
Pointer --> Health[Run health and retrieval checks]The full refresh uses staging, validation, publication, health checks, and rollback. A failed refresh does not change the Markdown authority.
Reliability and performance
The repository pins Graphify 0.9.26 in an isolated environment.
Scope filters run before provider ranking.
RRF combines independent provider rankings.
Bounded reranking limits query work.
Incremental indexing skips unchanged Markdown files.
Versioned snapshots preserve the last satisfactory index.
Full graph refreshes validate staged data before publication.
Evidence packets include canonical source paths.
Quick start
Install these items:
Git
Python 3.11 or later
PowerShell 5.1 or later
A Markdown memory directory
Open PowerShell in the repository root. Then, run this command:
.\scripts\setup.ps1 -MemoryRoot 'C:\path\to\AI-Memory' -InstallClientsRestart each configured client after the setup procedure is complete.
For more setup information, read the installation guide.
MCP tools
Tool | Function |
| Finds relevant memory evidence. |
| Gets one memory by its identity or path. |
| Gets related memories. |
| Finds a relationship path. |
| Explains why evidence matches a query. |
| Updates derived indexes. |
| Gives health and version data. |
| Records retrieval feedback. |
Repository layout
Path | Contents |
| MCP server, retrieval engine, indexer, and adapters |
| Setup, client installation, and Graphify operations |
| Canonical AI Memory skill |
| Automated behavior and portability tests |
| Frozen retrieval contract and fixtures |
| Architecture, setup, operations, and validation guides |
Documentation
The documentation index gives links to all project guides.
Source boundary
This is a public repository.
This repository contains all project source files.
It contains only neutral examples and synthetic benchmark fixtures.
The user memory directory stays outside Git.
Generated indexes, logs, and recovery files also stay outside Git.
Machine-specific and organization-specific values stay in the ignored .env file.
Read AGENTS.md before you change this repository.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/moorej2400/ai-memory-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server