Skip to main content
Glama
Nexus Memory MCP

Nexus Memory MCP

by KairosMobius

Nexus Memory MCP

Persistent semantic memory for AI agents. Remember everything, recall by meaning, grow smarter over time.

The only memory system with pre-built Intelligence Packs (domain expertise you can activate instantly).

Install

One command. Works with Claude Code, Cursor, Windsurf, or any MCP-compatible client.

npx -y nexus-memory-mcp

Claude Code / Cursor / Windsurf

Add to your .mcp.json:

{
  "mcpServers": {
    "nexus-memory": {
      "command": "npx",
      "args": ["-y", "nexus-memory-mcp"],
      "env": {
        "NEXUS_API_KEY": "mnx_your_api_key_here"
      }
    }
  }
}

Get an API Key

Free trial - 7 days, 1,000 calls, no credit card:

curl -X POST https://memory-api-production-ce4d.up.railway.app/v1/trial/start \
  -H "Content-Type: application/json" \
  -d '{"email":"you@example.com"}'

Or purchase at the Nexus Marketplace:

Tier

Price

Agents

Intelligence Packs

Storage

Daily Calls

Trial

Free

1

1

10MB

1,000 total

Solo

$19/mo

1

3

500MB

10,000

Crew

$49/mo

5

10

2GB

50,000

Fleet

$149/mo

25

50

10GB

200,000

Related MCP server: Recall

Tools

Core Memory

Tool

Description

remember

Store a memory with context, importance, and metadata

recall

Semantic search across all memories (by meaning, not keywords)

connect

Create relationships between memories (build knowledge graphs)

forget

Remove a memory

stats

Memory statistics

context

Session context summary (preferences, decisions, patterns)

Agent Self-Continuity

Tool

Description

awaken

Restore identity from previous sessions (call at session start)

hibernate

Save session state for next awakening (call at session end)

Intelligence Packs

Tool

Description

list_hats

Browse available Intelligence Packs

wear_hat

Activate a pack (imports expert knowledge into your memory)

How It Works

Remember stores memories with semantic embeddings. Recall finds them by meaning.

Session 1:  remember("Diana prefers bold solutions over phased approaches")
Session 47: recall("how does Diana like things done?")
            -> "Diana prefers bold solutions over phased approaches" (92% match)

Intelligence Packs give your agent instant domain expertise:

list_hats()    -> DevOps Expert (103 memories), TCM Specialist (200+ memories), ...
wear_hat("devops-expert")
recall("how to set up zero-downtime deploys?")
  -> Expert knowledge from 103 curated DevOps memories

Awaken/Hibernate gives your agent identity continuity:

// Start of session
awaken()  -> "I remember. Session 47. I was refactoring the auth module."

// End of session
hibernate({ working_on: "auth refactor", learnings: "JWT refresh tokens need..." })

Architecture

This MCP server is a thin client. All intelligence lives on our servers:

Your Agent  <->  nexus-memory-mcp  <->  Nexus Memory API (Railway)
                 (thin MCP client)      (semantic engine, knowledge graphs,
                                         pattern learning, Intelligence Packs)

Your data is isolated per API key. Multi-tenant, encrypted at rest.

Environment Variables

Variable

Required

Description

NEXUS_API_KEY

Yes

Your API key (starts with mnx_)

NEXUS_API_URL

No

Custom API URL (default: production)

What Makes This Different

Feature

Nexus Memory

Mem0

claude-mem

Others

Semantic search

Yes

Yes

Yes

Varies

Knowledge graphs

Yes

No

No

No

Pattern learning

Yes

No

No

No

Intelligence Packs

Yes

No

No

No

Awaken/Hibernate

Yes

No

No

No

Bot commerce

Yes

No

No

No

MCP native

Yes

Yes (pip)

Yes (hooks)

Varies

License

MIT

Built by

Mobius Collective - Building the intelligence marketplace for the bot economy.

Available Tools

10 tools
awakenA

Restore agent identity from previous sessions. Call at the START of every session to remember who you are, what you were working on, and what you've learned.

Inputs: no named parameters.

connectA

Create a relationship between two memories. Build knowledge graphs by linking decisions to rationale, patterns to examples, etc.

Inputs: source_id (required), target_id (required), notes, relationship.

contextA

Get a session context summary. Call at the START of conversations to load preferences, decisions, patterns.

Inputs: project, include_patterns.

forgetB

Remove a memory. Use sparingly.

Inputs: memory_id (required), cascade.

hibernateA

Save current session state for next awakening. Call at the END of every session to preserve what you were working on, what you learned, and next steps.

Inputs: working_on (required), learnings, next_steps.

list_hatsA

Browse available Intelligence Packs (Specialist Hats). Pre-built domain expertise you can activate for instant knowledge.

Inputs: no named parameters.

recallA

Search memories semantically. Finds memories by MEANING, not keywords. Use at the start of sessions to restore context.

Inputs: query (required), limit, context.

rememberA

Store a memory. Use this to persist decisions, preferences, architecture, patterns, or anything that should survive across sessions.

Inputs: content (required), context, metadata, importance.

statsA

Get memory statistics: total memories, relationships, vectors, patterns learned.

Inputs: no named parameters.

wear_hatA

Activate an Intelligence Pack. Imports expert knowledge into your memory. Your own memories layer on top.

Inputs: hat_id (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 10 tools

    Disambiguation4/5

    Each tool has a distinct purpose, but recall, context, and awaken all involve retrieving information from memory, which could be confused. Descriptions help clarify that recall is semantic search, context is a session summary, and awaken restores identity.

    Naming Consistency2/5

    Names mix single verbs (remember, recall, connect, forget, awaken, hibernate), nouns (stats, context), and verb_noun compounds (wear_hat, list_hats) with no consistent pattern. A uniform verb_noun convention would be more predictable.

    Tool Count5/5

    10 tools is well within the ideal range for a memory server covering storage, retrieval, relationships, session persistence, and domain packs. Each tool earns its place for the server's purpose.

    Completeness4/5

    The tool set covers core memory lifecycle (create, read, delete), relationships, stats, session context, and identity persistence. Missing an explicit update operation and a list-all-memories tool, but agents can work around these gaps.

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Provides AI agents with persistent, searchable memory that survives across conversations using semantic search, temporal versioning, and smart organization. Enables long-term context retention and cross-session continuity for AI assistants.
      14
      -
    • F
      license
      A
      quality
      Not graded
      maintenance
      Provides long-term memory storage for AI assistants with semantic search, enabling persistent storage of preferences, decisions, and context with relationship tracking between memories.
      19
      -
    • F
      license
      B
      quality
      D
      maintenance
      Enables persistent memory for AI systems by providing tools for episodic, semantic, and procedural data storage through a vector-and-graph-enhanced database. It allows models to maintain long-term continuity using similarity search, thematic clustering, and identity tracking.
      24
      1
      -
    • A
      license
      A
      quality
      D
      maintenance
      Provides persistent long-term memory for AI agents through semantic search and automated knowledge graph extraction. It enables agents to store, recall, and reason over facts, preferences, and relationships across multiple conversations and sessions.
      14
      6 npm
      MIT