RecallBricks Documentation

Memory Graph Infrastructure for AI Agents

RecallBricks gives your AI agents unlimited memory, pattern discovery, and cross-agent collaboration. This documentation covers the complete API and SDK reference.

Table of Contents

• Quick Start
• Installation
• Memories API
• LangChain Integration
• Relationships API
• Advanced Usage
• Migration Guide (v1.2)

---

Quick Start

Give your AI agent memory in 5 minutes.

1. Get Your API Key

Start your 14-day free trial (no credit card required) and generate your API key from the dashboard.

2. Install the SDK

Python:

pip install recallbricks

TypeScript:

npm install recallbricks

LangChain:

pip install recallbricks-langchain

3. Save Your First Memory

Python:

from recallbricks import RecallBricks

rb = RecallBricks(api_key="rb_your_api_key_here")

# Save a memory with automatic metadata extraction
result = rb.learn(text="User prefers dark mode")

# Auto-generated metadata:
print(result.metadata.tags)       # ['preferences', 'ui', 'settings']
print(result.metadata.category)   # 'Preferences'
print(result.metadata.importance) # 0.75

TypeScript:

import { RecallBricks } from 'recallbricks';

const rb = new RecallBricks({ apiKey: 'rb_your_api_key_here' });

// Save a memory with automatic metadata extraction
const result = await rb.learn({
  text: 'User prefers dark mode'
});

// Auto-generated metadata:
console.log(result.metadata.tags);       // ['preferences', 'ui', 'settings']
console.log(result.metadata.category);   // 'Preferences'
console.log(result.metadata.importance); // 0.75

cURL:

curl -X POST https://api.recallbricks.com/api/v1/memories/learn \
  -H "X-API-Key: rb_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{"text": "User prefers dark mode"}'

4. Recall Memories

Python:

# Standard recall
memories = rb.recall(query="user preferences", limit=10)

# Organized recall (NEW) - includes category summaries
result = rb.recall(
    query="user preferences",
    limit=10,
    organized=True
)

print(result.categories)  # Category summaries for faster agent reasoning
print(result.memories)    # Individual memories

TypeScript:

// Standard recall
const memories = await rb.recall({
  query: 'user preferences',
  limit: 10
});

// Organized recall (NEW) - includes category summaries
const organized = await rb.recall({
  query: 'user preferences',
  limit: 10,
  organized: true
});

console.log(organized.categories);  // Category summaries
console.log(organized.memories);    // Individual memories

cURL:

curl -X POST "https://api.recallbricks.com/api/v1/memories/recall" \
  -H "X-API-Key: rb_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{"query": "user preferences", "limit": 10, "organized": true}'

---

Installation

Python SDK

pip install recallbricks

GitHub: recallbricks/recallbricks-python

TypeScript SDK

npm install recallbricks

GitHub: recallbricks/recallbricks-ts

LangChain Integration

pip install recallbricks-langchain

GitHub: recallbricks/recallbricks-langchain

---

Authentication

All API requests require an API key. You can:

1. Pass it directly to the SDK:

   rb = RecallBricks(api_key="rb_your_api_key_here")

2. Use the RECALLBRICKS_API_KEY environment variable:

   export RECALLBRICKS_API_KEY="rb_your_api_key_here"

3. Use the X-API-Key header for REST API calls:

   curl -H "X-API-Key: rb_your_api_key_here" ...

---

Base URL

https://api.recallbricks.com/api/v1

---

What's New in v1.2

• learn() endpoint - Automatic metadata extraction (tags, categories, entities, importance)
• recall() with organized=true - Category summaries for 3-5x faster agent reasoning
• Memory decay - Penalize older memories with decay_old_memories=true
• Usage weighting - Boost frequently-used memories with weight_by_usage=true
• Deprecation: save() is deprecated in favor of learn()

See the Migration Guide for upgrade instructions.

---

Pricing

Pro Plan - $79/month

• 100,000 operations/month included
• Unlimited storage
• Automatic overages at $0.10 per 100 extra operations
• 14-day free trial (no credit card required)

What counts as an operation?

• Saving a memory (1 op)
• Recalling memories (1 op)
• Metadata extraction on first recall (1 op, then cached)

---

Rate Limits

Limit	Value
Per-second burst	100 requests/second
Per-minute heavy burst	3,000 requests/minute
Monthly operations	100,000 included (Pro)

---

Support

• Email: support@recallbricks.com
• GitHub Issues: recallbricks/recallbricks-python/issues

---

Next Steps

• Memories API Reference - Full CRUD operations
• LangChain Integration - Persistent memory for LangChain
• Relationships API - Explore memory connections
• Advanced Usage - Metadata extraction, decay, raw mode