RushDB supports automatic runtime memory, model-directed tools, and application-controlled memory without forcing them into one abstraction.

Native memory owns automatic recall and completed-turn capture. MCP remains the explicit knowledge plane for records such as DECISION, TASK, ENTITY, and ARTIFACT, so the same turn is not written twice.

openclaw plugins install @rushdb/openclaw-memory
export RUSHDB_API_KEY="your-project-api-key"

The additive plugin recalls before prompt construction and durably captures successful completed turns while leaving OpenClaw's Markdown and SQLite memory active.

OpenClaw setup and behavior →

pip install rushdb-hermes-memory
export RUSHDB_API_KEY="your-project-api-key"
hermes memory setup

The native MemoryProvider participates in prefetch, completed-turn sync, pre-compression, session changes, session end, and shutdown. Supported explicit Hermes memory writes become scoped memory facts.

Hermes setup and behavior →

Both connectors:

• apply trusted agent, profile, privacy, participant, and sandbox scope before semantic ranking;
• capture bounded user/assistant content instead of complete transcripts or tool output;
• fail open when remote recall is slow or unavailable;
• write through a local durable outbox and replay after restart;
• use deterministic identities and idempotent RushDB upserts;
• keep a recent-write fallback while managed embeddings become visible.

AgentMemoryEvent v1 gives native connectors and custom harnesses the same provider-neutral memory model.

The @rushdb/agent-memory-contract package publishes TypeScript primitives, deterministic identity helpers, the JSON Schema, and a language-neutral conformance fixture.

pnpm add @rushdb/agent-memory-contract

For ChatGPT, Claude.ai, and other remote clients, connect directly to:

https://mcp.rushdb.com/mcp

For local MCP clients such as Claude Desktop, Cursor, Windsurf, and VS Code:

{
  "mcpServers": {
    "rushdb": {
      "command": "npx",
      "args": ["@rushdb/mcp-server"],
      "env": {
        "RUSHDB_API_KEY": "your-project-api-key"
      }
    }
  }
}

MCP installation and OAuth setup →

Get an API key at app.rushdb.com, then install a RushDB SDK:

pnpm add @rushdb/javascript-sdk
# or
pip install rushdb

import RushDB from '@rushdb/javascript-sdk'

const db = new RushDB('RUSHDB_API_KEY')

await db.ai.indexes.create({
  label: 'MEMORY',
  propertyName: 'text'
})

await db.records.create({
  label: 'MEMORY',
  data: {
    agentId: 'agent-42',
    topic: 'authentication',
    text: 'Use passkeys for the new account flow.'
  }
})

const memories = await db.records.vectorSearch({
  labels: ['MEMORY'],
  propertyName: 'text',
  query: 'What did we decide about authentication?',
  where: { agentId: 'agent-42' },
  limit: 10
})

RushDB manages embeddings server-side. The same query can combine semantic ranking with structured filters and graph traversal.

You also get filtering, aggregation, transactions, CSV and JSON import, cloud and self-hosted deployment, TypeScript and Python SDKs, REST, MCP, and agent skills.