MindCache is a TypeScript library for managing short-term memory in AI agents through a simple, LLM-friendly key-value repository. Perfect for chatbots, AI assistants, and intelligent workflows.
Built specifically for AI agents and LLM integration
Automatically generate LLM tool schemas for AI agents to read and write memory with zero configuration
Intelligent context tracking and memory organization that adapts to your AI agent's needs
Dynamically generate system prompts with all memory context and tool instructions included
Dynamic {placeholder} replacement with recursive resolution for context-aware prompts
Subscribe to specific keys or all changes with event-driven listeners
Organize memory with tags and search by category for efficient retrieval
Real-world use cases for AI agents
// Auto Tool Generation for AI Agents
import { mindcache } from 'mindcache';
import { generateText } from 'ai';
// Store user preferences and context
mindcache.set_value('userName', 'Alice', { tags: ['user'] });
mindcache.set_value('favoriteColor', 'blue', { tags: ['preferences'] });
mindcache.set_value('lastTask', 'planning vacation', { tags: ['context'] });
// Automatically generate tool schemas for Vercel AI SDK
const tools = mindcache.get_aisdk_tools();
// Returns: {
// write_userName: { description: "...", parameters: {...} },
// write_favoriteColor: { description: "...", parameters: {...} },
// write_lastTask: { description: "...", parameters: {...} }
// }
// Generate system prompt with memory context
const systemPrompt = mindcache.get_system_prompt();
// "userName: Alice. You can rewrite \"userName\" by using..."
// "favoriteColor: blue. You can rewrite \"favoriteColor\" by..."
// "$date: 2025-01-15"
// Use with AI SDK - tools are automatically available
const { text, toolCalls } = await generateText({
model: openai('gpt-4'),
tools: tools, // AI can now call write_userName(), etc.
system: systemPrompt,
prompt: 'Remember that I love green now, not blue.'
});
// AI will automatically call write_favoriteColor('green')
// and the memory is updated automaticallyAutomatically generate LLM tool schemas so AI agents can read and write memory using function calls. No manual tool definitions needed - tools are created for every writable key in your memory.
// Context Management with Tags & System Prompts
import { mindcache } from 'mindcache';
// Mark context items with tags
mindcache.set_value('userName', 'Alice', { tags: ['context'] });
mindcache.set_value('userRole', 'developer', { tags: ['context'] });
mindcache.set_value('projectName', 'MindCache', { tags: ['context'] });
// Store non-context data without tags
mindcache.set_value('tempNote', 'Meeting at 3pm');
mindcache.set_value('cacheVersion', '1.0.0');
// Extract only tagged context for focused prompts
const contextData = mindcache.getTagged('context');
// "userName: Alice, userRole: developer, projectName: MindCache"
// Generate system prompt with all visible keys
const fullSystemPrompt = mindcache.get_system_prompt();
// Includes all keys: userName, userRole, projectName, tempNote, cacheVersion
// Or create focused context from tagged items only
const focusedContext = `Context: ${contextData}
Instructions: You are helping ${mindcache.get_value('userName')}
with their ${mindcache.get_value('projectName')} project.`;
// Use focused context for specific workflows
// Or use full system prompt for comprehensive memory access
// Filter keys by tag programmatically
const allKeys = Object.keys(mindcache.getAll());
const contextKeys = allKeys.filter(key =>
mindcache.hasTag(key, 'context')
);
// ['userName', 'userRole', 'projectName']Use tags to mark important context items, extract only tagged content when needed, and automatically generate system prompts with all memory context. Perfect for building focused AI workflows while maintaining full memory access.