@memsolus/mcp

MCP Server for Memsolus — persistent memory, knowledge graphs, and semantic search for AI agents

Documentation  |  npm  |  GitHub  |  memsolus.com

---

What is Memsolus?

Memsolus is a persistent memory platform for AI agents. It lets your AI assistant remember context across conversations, track decisions, search past knowledge, and navigate entity relationships through a knowledge graph.

This package exposes the full Memsolus API as an MCP (Model Context Protocol) server, so any MCP-compatible AI client — Claude Desktop, Claude Code, Cursor, Windsurf, VS Code Copilot, Zed, and others — can store and retrieve memories, query knowledge, and traverse entity graphs without custom integration code.

---

Prerequisites

• A Memsolus account (sign up)
• An API key (see Getting an API Key below)
• Node.js 18+ or Bun

---

Getting an API Key

1. Sign up or log in at app.memsolus.com
2. Go to Settings > API Keys
3. Click Create API Key
4. Give it a name (e.g., Claude Desktop)
5. Select permissions: at minimum MemoryRead and MemoryWrite. For full access, also enable KnowledgeRead and DashboardRead
6. Copy the key — it starts with msk_ and you will not be able to see it again after closing the dialog
7. Set the key as the MEMSOLUS_API_KEY environment variable or paste it directly into your client config

---

Installation

Choose the section that matches your AI client. All configurations use npx -y @memsolus/mcp, which runs the latest published version without a local install step.

Claude Code

Run this command in your terminal:

claude mcp add memsolus -- npx -y @memsolus/mcp

To pass the API key inline without setting an environment variable:

claude mcp add memsolus -e MEMSOLUS_API_KEY=msk_your_key_here -- npx -y @memsolus/mcp

After adding, restart Claude Code or run /mcp to confirm that memsolus appears in the server list.

---

Claude Desktop

Locate and open your Claude Desktop configuration file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the following entry inside mcpServers:

{
  "mcpServers": {
    "memsolus": {
      "command": "npx",
      "args": ["-y", "@memsolus/mcp"],
      "env": {
        "MEMSOLUS_API_KEY": "msk_your_key_here"
      }
    }
  }
}

Save the file and restart Claude Desktop. The Memsolus tools will appear in the tools panel on the next launch.

---

Cursor

Create or edit .cursor/mcp.json in your project root (or the global Cursor MCP config at ~/.cursor/mcp.json for workspace-independent access):

{
  "mcpServers": {
    "memsolus": {
      "command": "npx",
      "args": ["-y", "@memsolus/mcp"],
      "env": {
        "MEMSOLUS_API_KEY": "msk_your_key_here"
      }
    }
  }
}

Reload Cursor or run the MCP: Restart Servers command from the command palette.

---

Windsurf

Open your Windsurf MCP configuration (~/.codeium/windsurf/mcp_config.json) and add:

{
  "mcpServers": {
    "memsolus": {
      "command": "npx",
      "args": ["-y", "@memsolus/mcp"],
      "env": {
        "MEMSOLUS_API_KEY": "msk_your_key_here"
      }
    }
  }
}

Restart Windsurf to apply the changes.

---

VS Code (GitHub Copilot)

Add a .vscode/mcp.json file to your workspace:

{
  "mcpServers": {
    "memsolus": {
      "command": "npx",
      "args": ["-y", "@memsolus/mcp"],
      "env": {
        "MEMSOLUS_API_KEY": "msk_your_key_here"
      }
    }
  }
}

Alternatively, add the same block to your VS Code user settings.json under "mcp" to enable it globally. Reload the VS Code window after saving.

---

Zed

Open your Zed settings (~/.config/zed/settings.json) and add a context_servers entry:

{
  "context_servers": {
    "memsolus": {
      "command": {
        "path": "npx",
        "args": ["-y", "@memsolus/mcp"],
        "env": {
          "MEMSOLUS_API_KEY": "msk_your_key_here"
        }
      }
    }
  }
}

Restart Zed to pick up the new server.

---

Using an Environment Variable Instead

If you prefer not to embed your API key in config files, export it in your shell profile (~/.bashrc, ~/.zshrc, or equivalent):

export MEMSOLUS_API_KEY=msk_your_key_here

With the variable set globally, all client configs can omit the env block:

{
  "mcpServers": {
    "memsolus": {
      "command": "npx",
      "args": ["-y", "@memsolus/mcp"]
    }
  }
}

---

Configuration

The server is configured entirely through environment variables. No config file is needed.

---

Available Tools

The server exposes 17 tools across four categories. Which tools are available depends on your plan — see Plan-Based Tool Availability.

Memory Tools

These tools are available on all plans that include MCP access.

add_memory

Store a new persistent memory. Use when the user shares a fact, preference, decision, or any context worth remembering across future conversations.

---

search_memories

Search stored memories by relevance. Supports semantic, keyword, and hybrid search modes.

Use hybrid for general queries. Use keyword when searching for exact names, IDs, or technical terms.

---

get_memories

List memories in chronological order. Use when you want to browse all stored memories rather than search by relevance.

---

get_memory

Retrieve a single memory by its ID.

---

update_memory

Update the content or priority of an existing memory. Prefer this over creating a duplicate when the user corrects or refines information that is already stored.

---

delete_memory

Permanently delete a memory. Use when the user explicitly asks to forget something, or when a memory is clearly outdated. Search first if you do not have the ID.

---

promote_memory

Promote a temporary memory to permanent status. Memories stored as task-scoped (with an expiration) can be made permanent with this tool when the context proves to be persistently relevant.

---

submit_feedback

Submit quality feedback on a memory. Positive and negative signals improve memory quality over time.

---

get_memory_history

Retrieve the full event history of a memory: creation, updates, promotions, and feedback signals. Use to understand how a memory evolved or diagnose unexpected behavior.

---

Knowledge Tools

These tools require a plan that includes Knowledge Graph access (Pro or above).

get_knowledge

Get the consolidated knowledge profile built from processed memories. Set merged=true (the default) to receive a complete Markdown profile. Set merged=false to paginate through individual knowledge entries.

Call this at the start of a conversation to load the full context profile for a user before responding.

---

get_knowledge_entry

Retrieve a specific knowledge entry by ID, including the source memory IDs that contributed to it.

---

Graph Tools

These tools require a plan that includes Knowledge Graph access (Pro or above).

graph_search

Search for entities in the knowledge graph. Entities include people, organizations, places, and things mentioned across memories.

---

graph_traverse

Navigate relationships starting from a known entity. Useful for exploring who an entity is connected to — colleagues, projects, organizations. Use graph_search first if you need to find the entity ID.

---

graph_query

Answer a natural language question using the knowledge graph. Returns an answer with a confidence score and the entities used to construct it.

---

Utility Tools

These tools are available on all plans that include MCP access.

get_dashboard_summary

Get workspace KPIs — total memories stored, searches performed, and tokens consumed — with a comparison to the previous period.

---

get_memory_profile

Get a summarized profile of a user built from their stored memories: key topics, entity count, and a narrative summary.

---

list_entities

List all users and agents in the workspace with their memory counts. Use to discover who has memories stored or to compare activity across entities.

This tool takes no parameters.

---

Resources

Resources are read-only data endpoints that MCP clients can load as context. They are accessible via URI.

memsolus://knowledge/merged is only available on plans with Knowledge Graph access.

---

Prompts

Prompts are pre-built interaction templates that combine tool calls with structured output. Clients that support MCP prompts can invoke them by name.

recall-context

Retrieve and format relevant context before starting a task. Searches memories and loads the knowledge profile for a given topic, returning both in a single response.

---

session-handoff

Save relevant context at the end of a session. Stores the provided summary as a high-priority persistent memory so it is available in future conversations.

---

decision-log

Record a decision as a high-priority persistent memory, including the alternatives that were considered and the reasoning behind the final choice.

---

Plan-Based Tool Availability

The server checks your plan entitlements at startup and registers only the tools your plan includes. This is done automatically — you do not need to configure anything.

If a tool does not appear in your client’s tool list, it is because your current plan does not include that feature. Upgrade at app.memsolus.com/settings/billing.

If the server cannot reach the Memsolus API at startup to verify your entitlements, it registers all tools by default and lets the API enforce access. This prevents startup failures caused by temporary network issues.

---

Examples

Storing a decision

Tell your assistant:

“Remember that we decided to use PostgreSQL instead of MongoDB for the user service. The reason was strong consistency requirements for financial transactions.”

The assistant calls add_memory with priority HIGH, and the decision is stored as a permanent memory associated with your workspace.

---

Searching past context

Ask your assistant:

“What do you remember about our authentication architecture?”

The assistant calls search_memories with query "authentication architecture" in hybrid mode, retrieves ranked results, and summarizes what Memsolus knows on that topic.

---

Exploring the knowledge graph

Ask your assistant:

“Who are the main people involved in the billing module and how are they connected?”

The assistant calls graph_search to find entities related to the billing module, then calls graph_traverse on the most relevant entity to map its relationships. It can also use graph_query directly:

“What projects is the backend team involved in?”

graph_query takes the natural language question, answers it using the knowledge graph, and returns the result with a confidence score.

---

Loading context at the start of a conversation

For clients that support MCP prompts, invoke recall-context with the topic you are about to work on:

prompt: recall-context
arguments:
  topic: "billing refactor"

This runs a memory search and loads the full knowledge profile in a single step, giving the assistant the full context before it responds.

---

Saving context at the end of a session

Invoke session-handoff before closing the conversation:

prompt: session-handoff
arguments:
  summary: "Reviewed billing module. Decided to split the invoice service into two microservices. Deferred payment retry logic to next sprint."

The summary is stored as a high-priority memory and will be available when you return.

---

Troubleshooting

Tools do not appear in my client

1. Confirm that MEMSOLUS_API_KEY is set and starts with msk_
2. Confirm that your plan includes MCP access — check app.memsolus.com/settings/billing
3. Restart your MCP client after changing the config
4. On Claude Desktop, check ~/Library/Logs/Claude/ for server startup errors

Knowledge and graph tools are missing

These tools require a Pro plan or above. If you are on a free or starter plan, only memory and utility tools are registered. Upgrade at app.memsolus.com/settings/billing.

UNAUTHORIZED error when calling a tool

Your API key is invalid or has expired. Generate a new key at app.memsolus.com/settings/api-keys and update your config.

ENTITLEMENT_NOT_AVAILABLE error

The tool or feature you are trying to use is not included in your plan. The server registered the tool because entitlements could not be confirmed at startup, but the API rejected the call. Upgrade your plan or check that your API key belongs to the correct workspace.

RATE_LIMIT error

You have exceeded the request rate for your plan. Wait a few seconds and retry. If you are hitting rate limits consistently, consider batching calls or upgrading to a plan with higher limits.

The server fails to start with MEMSOLUS_API_KEY environment variable is required

The environment variable is not visible to the MCP server process. Make sure the env block in your client config includes the key explicitly, or that the variable is exported in the shell environment that launches your client.

---

Related Packages

• TypeScript SDK: @memsolus/sdk — programmatic access to the Memsolus API
• Claude Code Plugin: memsolus/claude-plugin — deeper Claude Code integration
• Documentation: docs.memsolus.com
• Website: memsolus.com

---

License

MIT