Skip to content
MCP server

Semurg agent memory

Self-hosted persistent memory for agents: knowledge graph, vectors and cited recall

Runs locallyStatus and history
Pricing
Free
Platforms
self-hosted, Linux, Windows
Runs
Locally
Version
11.153.20261103
Listing updated
3 days ago

About Semurg agent memory

Semurg agent memory is a free MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Semurg gives an agent a memory that survives the end of a session, running on your own machine on a single engine the authors wrote from scratch. The node sends nothing to a hosted service, and one node is free with no licence key. It runs on x86_64 hardware: Ubuntu 24.04, Windows through WSL2, or Docker.

Recall is built around citations. Each returned item says which document it came from and when it was stored, and for passages and facts read from text it carries the quoted sentence, checked against the stored text before it is served; if a quote cannot be verified the span is left empty and the reply says why. Facts an agent sent without text are labelled as asserted. When nothing matches, the reply says nothing_stored.

Memories are separated by scope, and an app can give each end user their own with user_id, pinned by a header. Forgetting an entity hides it and its linked facts and passages immediately. Semantic search works once a local embedding model is installed.

Install

Run in a terminal

claude mcp add agent-memory --env SEMURG_API_TOKEN='<SEMURG_API_TOKEN>' -- uvx semurg

Add to claude_desktop_config.json

{
  "mcpServers": {
    "agent-memory": {
      "command": "uvx",
      "args": [
        "semurg"
      ],
      "env": {
        "SEMURG_API_TOKEN": "<SEMURG_API_TOKEN>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "agent-memory": {
      "command": "uvx",
      "args": [
        "semurg"
      ],
      "env": {
        "SEMURG_API_TOKEN": "<SEMURG_API_TOKEN>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "agent-memory": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "semurg"
      ],
      "env": {
        "SEMURG_API_TOKEN": "<SEMURG_API_TOKEN>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "agent-memory": {
      "command": "uvx",
      "args": [
        "semurg"
      ],
      "env": {
        "SEMURG_API_TOKEN": "<SEMURG_API_TOKEN>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "agent-memory": {
      "command": "uvx",
      "args": [
        "semurg"
      ],
      "env": {
        "SEMURG_API_TOKEN": "<SEMURG_API_TOKEN>"
      }
    }
  }
}

Configuration

  • SEMURG_URLYour Semurg node (http://127.0.0.1:4000 by default)
  • SEMURG_API_TOKENYour node API tokenRequired
  • SEMURG_SCOPEA default memory scope for this connection

Tools this server provides (9)

Taken from the project documentation; the running server may differ.

memory_status
Reports memory status.
memory_remember
Stores what the agent learns.
memory_recall
Finds stored items again.
memory_relate
Creates relationships between entities.
memory_forget
Hides an entity, its facts and linked passages.
memory_context
Starts from names in a question and returns related facts and passages.
memory_document
Works with documents.
memory_export
Reads everything a scope holds, a page at a time.
memory_import
Stores records back into any scope.

What it does

Knowledge graph
Entities and relationships recorded without a schema to write.
Cited answers
Items cite their source document, with quotes checked against stored text.
Question-based context
memory_context walks relationships around names found in a question.
Scoped memories
Share or separate memory by scope and per-user ids.
Immediate forget
Forgetting hides an entity and its linked passages at once.
Local semantic search
Meaning-based search runs on your node once an embedding model is installed.

Good for

  • Give a coding agent memory across sessions
  • Keep per-user memory in an app
  • Export and import a memory scope
  • Answer questions with citations to stored sources
  • Keep agent memory off hosted services

Pricing

Free

One node is free with no licence key

Prices change. Check one.semurg.io before you buy.

Semurg agent memory FAQ

Does data leave my machine?

The post says the node sends nothing to a hosted service.

What does it cost?

One node is free, with no licence key.

What hardware is needed?

x86_64: Ubuntu 24.04, Windows via WSL2 with Ubuntu 24.04, or Docker on an x86_64 host.

What happens when nothing matches?

The reply says nothing_stored.

Is Semurg agent memory free?

Yes. Semurg agent memory is free to use.

How do I add Semurg agent memory to Claude Code?

Run this command in your terminal: claude mcp add agent-memory --env SEMURG_API_TOKEN='<SEMURG_API_TOKEN>' -- uvx semurg

What are the best alternatives to Semurg agent memory?

The closest alternatives to Semurg agent memory are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.

Alternatives to Semurg agent memory

  1. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  2. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  3. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  4. Local-first long-term memory layer for AI agents, served over MCP

    Not checked yet
    • 5 tools
    • Remote
    • Open source
    • 1.3K GitHub stars
    • Knowledge and memory
  5. note1

    MCP server

    MCP server giving AI assistants search and summaries across your note1 meetings

    • 4 tools
    • Remote
    • Free trial
    • 1 GitHub star
    • Productivity