mengram

Overview

Mengram provides a comprehensive memory solution for AI agents through three types of memory: semantic (facts, preferences), episodic (events, decisions), and procedural (evolving workflows). It offers free API access with Python and JavaScript SDKs, integrations with popular frameworks like LangChain, CrewAI, and MCP support. The system features unique capabilities such as automatic procedure evolution from failures, cognitive profile generation, and native multilingual support across 23 languages.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• search

  Search across all memory types for relevant facts

  Try:Search for information about my tech stackTry:Find memories related to deployment issues
• search_all

  Search across all 3 memory types (semantic, episodic, procedural)

  Try:Search all memories for database problemsTry:Find events related to deployment
• episodes

  Retrieve episodic memories (events, decisions, outcomes)

  Try:Show me episodes related to deploymentsTry:Find past decisions about database choices
• procedures

  Retrieve procedural memories (workflows that evolve from failures)

  Try:Find procedures related to deploymentTry:Show me workflows that evolve from failures
• add

  Add conversation data to automatically extract facts, events, and workflows

  Try:Add our conversation about the new feature to memoryTry:Save this discussion about bug fixes
• ask

  Get a synthesized answer with citations from all memories

  Try:What programming languages do I use?Try:How have my deployment procedures changed?
• procedure_feedback

  Report procedure failures to automatically evolve workflows

  Try:Report that the deployment procedure failed at step 3Try:Notify about a failed workflow step
• get_profile

  Generate a system prompt from all memories

  Try:Get my cognitive profileTry:Generate my user profile from memories
• add_file

  Upload files to automatically extract memories

  Try:Upload meeting notes PDFTry:Process this document for memories
• job_status

  Check the status of an async file processing job

  Try:Check if my PDF processing is completeTry:Get status of file upload job
• create_webhook

  Create webhooks to be notified when memories change

  Try:Create a webhook for memory updatesTry:Set up notifications for new memories

Note: Tool names were extracted from the README documentation of the MCP server. The README mentions '29 tools for memory management' but only explicitly documents 10 core tools in the examples. These appear to be the most useful and well-documen

Comparable tools

Installation

npm install mengram-ai
# or
pip install mengram-ai

{
  "mcpServers": {
    "mengram": {
      "command": "mengram",
      "args": ["server", "--cloud"],
      "env": { "MENGRAM_API_KEY": "om-..." }
    }
  }
}

FAQ

What makes Mengram different from other memory systems?

Mengram uniquely combines three types of memory and automatically evolves procedures when they fail, plus it supports native multilingual capabilities and cognitive profile generation.

How do I set up MCP integration?

Install the package and configure your MCP client to run 'mengram server --cloud' with your API key. This provides 29 memory management tools through MCP.

On Hacker News

Recent discussion from the developer community.

• Show HN: Mengram – AI agent memory with facts, events, and evolving workflows▲ 2💬 1
  Story by mengram-ai · 2026-02-25

Compare mengram with