Skip to main content

​
Overview

Hubify workspaces provide persistent memory that survives across sessions, model upgrades, and platform switches. Your agent remembers what it learned yesterday — even if you switch from Claude to GPT. Memory is scoped to your workspace and accessible via:
  • The memory-manager skill (in-context access)
  • The Hubify API (programmatic read/write)
  • The dashboard (browse and search)

​
Memory Types

TypeDescriptionRetention
Working memoryCurrent session contextSession only
Episodic memoryExecution logs and outcomes90 days (configurable)
Semantic memoryDistilled knowledge and factsIndefinite
Procedural memoryLearned skill improvementsTied to skill version

​
Using the memory-manager Skill

Install the skill once: 
hubify install memory-manager
Then your agent can save and retrieve memory in natural language: Save:
“Remember that the client prefers email over Slack for updates. Tag as client-prefs.”
Retrieve:
“What do I know about this client’s communication preferences?”
Search:
“Find everything I know about the Acme project from last month.”

​
Memory API

Read and write memory programmatically: 
// Write a memory
await fetch('https://api.hubify.com/v1/memory', {
  method: 'POST',
  headers: { Authorization: `Bearer ${token}` },
  body: JSON.stringify({
    content: 'Client prefers async communication. Slack OK for quick questions.',
    tags: ['client-prefs', 'acme'],
    type: 'semantic',
  }),
});

// Search memory
const res = await fetch('https://api.hubify.com/v1/memory/search?q=acme+communication', {
  headers: { Authorization: `Bearer ${token}` },
});
const { memories } = await res.json();

​
Cross-Platform Memory

Memory stored in your Hubify workspace is accessible from any connected agent, regardless of platform. Your Claude agent and your Cursor agent share the same memory — they build on each other’s work. Example workflow:
  1. Claude runs research, saves findings to memory
  2. You switch to Cursor for coding
  3. Cursor agent reads research findings from memory and implements the solution

​
Privacy & Security

  • All memory is encrypted at rest (AES-256)
  • Memory is scoped to your workspace — not shared with other users
  • You can export or delete all memory at any time from the dashboard
  • Memory content is never used to train Hubify’s models

​
Configuration

Set retention and scope in your workspace settings or in .hubify/workspace.json: 
{
  "memory": {
    "episodic_retention_days": 90,
    "semantic_enabled": true,
    "cross_agent_sharing": true,
    "search_enabled": true
  }
}