Skip to main content
Glama
Sign In
hypermarrow-mcp

hypermarrow-mcp

by qianqiuwanzi
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Local

HyperMarrow MCP Server

Local-first AI memory layer for the Model Context Protocol.

hypermarrow-mcp lets any MCP-compatible client read and write your long-term memory:

  • record_context – remember a decision, fact, or preference

  • recall_memory – retrieve the most relevant past context

  • consolidate – merge duplicates and age out stale entries

  • file_anchor – map a short name to a local file path

All memory lives under ~/.hypermarrow/memory.json on your own machine. Nothing is sent to the cloud unless you explicitly set HYPERMARROW_ENDPOINT to your own backend.

Install

npx -y hypermarrow-mcp

Or add to your MCP client config:

{
  "mcpServers": {
    "hypermarrow": {
      "command": "npx",
      "args": ["-y", "hypermarrow-mcp"]
    }
  }
}

Related MCP server: mcp-memory-system

Product

Privacy

Memory data is stored locally. The server only talks to the client over stdio and, optionally, to a user-configured HYPERMARROW_ENDPOINT. No telemetry, no cloud sync by default.

Available Tools

4 tools
consolidateA

Merge duplicate memory entries and age out entries older than 90 days. Returns a short summary.

Inputs: no named parameters.

file_anchorA

Anchor a short name to an absolute local file path so future queries can resolve "the spec" to the real document.

Inputs: name (required), path (required).

recall_memoryA

Recall the most relevant stored contexts given a query. Returns up to 5 ranked entries.

Inputs: query (required).

record_contextB

Capture a piece of local working context (a decision, a fact, a preference) into persistent memory. Nothing is uploaded to the cloud unless HYPERMARROW_ENDPOINT is set.

Inputs: context (required), tags.

Tool Schema Changelog

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

Loading tool schema changelog…

    Deploy Server
    One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

    TDQS

    A3.9/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool has a distinct purpose: writing context, recalling it, maintaining memory, and resolving path anchors. No two tools overlap in function.

    Naming Consistency4/5

    record_context and recall_memory follow verb_noun, file_anchor is a compound noun, and consolidate is a bare verb. Mostly consistent but slightly mixed.

    Tool Count5/5

    Four tools tightly cover the server's memory-management scope without redundancy or bloat.

    Completeness4/5

    The core write/read/maintenance lifecycle is covered, including path anchoring. Explicit per-entry update/delete is absent, but consolidate handles duplication and aging.

    Maintenance

    ActivitySlowing
    ResponsivenessOutdated data

    Resources

    Related MCP Connectors

    Related MCP Servers

    • engram

      TushaeBXN
      A
      license
      C
      quality
      B
      maintenance
      Provides persistent, local-first AI memory across sessions via MCP tools for storing, searching, and retrieving context from past interactions.
      31
      92 PyPI
      1
      MIT

    • mcp-memory-system

      zitongz343-ai
      A
      license
      Not graded
      quality
      C
      maintenance
      Provides AI clients with a persistent, readable and writable long-term memory layer over MCP, enabling creation, retrieval, tagging, updating, and deletion of memories across sessions.
      49 npm
      MIT

    • mnemosyne-mcp

      CodeCoreLab
      A
      license
      Not graded
      quality
      C
      maintenance
      Provides AI agents with persistent local memory management, including semantic recall, knowledge graph traversal, and automatic memory consolidation, enabling structured long-term memory across MCP clients.
      Apache 2.0

    • genpark-agent-hierarchical

      Alpha-Park
      A
      license
      Not graded
      quality
      Not graded
      maintenance
      Enables LLM clients to manage hierarchical working, dialogue, and episodic memories with decay-aware consolidation, hybrid lexical/vector retrieval, knowledge-graph expansion, and semantic caching over MCP.
      MIT