Memory infrastructure
for autonomous AI fleets.

from memorybrain import MemoryBrain

mb = MemoryBrain(api_key="mb_live_a1b2c3d4e5f6...")

# 1. Asynchronously store conversational facts & preferences (< 10ms)
mb.store(user_id="user_123", content="User prefers dark mode, communicates via Slack, deployed in AWS us-east-1.")

# 2. Sub-25ms hybrid vector (pgvector HNSW) + BM25 keyword recall
context = mb.recall(user_id="user_123", query="Where is the user's infrastructure hosted?")

print(context["context"])
# Output: "• Deployed in AWS us-east-1 • Communicates via Slack"

import { MemoryBrain } from "@memorybrain/sdk";

const mb = new MemoryBrain({ apiKey: "mb_live_..." });

// 1. Store turn-by-turn agent decision
await mb.store({
  userId: "enterprise_agent",
  content: "Approved budget cap of $50,000 for Q3 marketing automation.",
  category: "DECISION"
});

// 2. Recall active decision bounds
const memories = await mb.recall({
  userId: "enterprise_agent",
  query: "What is the Q3 marketing budget?"
});

// Claude Desktop / Cursor Config (claude_desktop_config.json)
{
  "mcpServers": {
    "memorybrain": {
      "command": "npx",
      "args": ["-y", "@memorybrain/mcp-server"],
      "env": {
        "MEMORYBRAIN_API_KEY": "mb_live_..."
      }
    }
  }
}
// 8 Native Tools: recall_memories, save_memory, search_data_assets,
// get_table_schema, get_metric_definition, get_data_lineage, etc.

# Pillar 26: Certified Enterprise Data Warehouse & Metric Context
hits = mb.assets.recall(
    query="Which table in our 5,000-table Snowflake lakehouse has verified churn metrics?",
    platform="snowflake",
    certified_only=True
)

# Returns in 3.1ms:
# - FQN: snowflake.analytics_prod.core.dim_customer_churn
# - Certified by: VP of Finance & Analytics
# - Health: HEALTHY (Alerts if upstream ETL pipeline failed)
# - Primary Table Columns: [customer_id, churn_date, lost_mrr, ...]
# - Canonical Join: JOIN analytics_prod.core.dim_customers ON ...