+15.33pp on SWE-bench Lite n=150 · p<0.05

The continuity layer
for intelligence.

Memory infrastructure that compounds across sessions, models, and agents. store, recall, dream — same shape across Python, Node, CLI, and MCP.

Get an API key Read the docs See the CLI

+15.33pp SWE-bench Lite · Opus + REM · paired bootstrap

@remlabs/sdk

// Five calls. Same shape across Python, Node, CLI, MCP.
import { rem } from '@remlabs/sdk';

await rem.store({
  key:       'pref_drink',
  value:     'user prefers cold-brew, oat milk',
  namespace: 'agent_42',
});

const hits = await rem.recall({
  query:     'morning routine',
  namespace: 'agent_42',
});

await rem.dream({ strategies: ['consolidate'] });

CLI

CLI · private beta — request access in Discord

Dump data with your email. Claim it later. Recall from any terminal. @remlabs/cli is in private beta — the npm package is gated until launch.

Terminal — preview (private beta)

# Preview only -- the @remlabs/cli npm package is in private beta.
# Request access: https://discord.gg/ux8NYVfK2

# 1. Dump — prompts for email, creates provisional account
$ npx @remlabs/cli dump chatgpt ~/Downloads/chatgpt-export.zip
#   Email: you@example.com
#   847 memories stored.
#   Sign in at remlabs.ai/claim to keep. Deleted after 4 days.

# 2. Sign in at remlabs.ai/claim with that email, then:
$ npx @remlabs/cli login

# 3. Recall — search your memories
$ npx @remlabs/cli recall "contact preferences"

# 4. Pipe to tools
$ npx @remlabs/cli recall "auth flow" --json | jq '.[] | .content'

# 5. Dream — run consolidation
$ npx @remlabs/cli dream --strategy synthesize

# 6. Compose — recall feeds into dream
$ npx @remlabs/cli recall "customer data" --json | npx @remlabs/cli dump --stdin

Supports ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, plus any JSON, CSV, Markdown, or plain text. Format auto-detected. See all platforms →

Import your data Claim with code

SDK Quickstart

SDK quickstart

Python, Node.js, or cURL. Five steps.

        
        
        
      

      # 1. Install
pip install remlabs

# 2. Initialize
from remlabs import RemLabs
rem = RemLabs("your-api-key")

# 3. Store
rem.store(key="user_pref", value="Prefers dark mode")

# 4. Recall (natural-language query)
results = rem.recall(query="user preferences")

# 5. List or fetch by key
items = rem.list(namespace="default", limit=10)
    

      // 1. Install -- private beta, request access in Discord
npm install @remlabs/sdk

// 2. Initialize
import { RemLabs } from '@remlabs/sdk';
const rem = new RemLabs('your-api-key');

// 3. Store
await rem.store({ key: 'user_pref', value: 'Prefers dark mode' });

// 4. Recall (natural-language query)
const results = await rem.recall({ query: 'user preferences' });

// 5. List
const items = await rem.list({ namespace: 'default', limit: 10 });
    

      # 1. Store a memory
curl -X POST https://remlabs.ai/v1/memory \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"key":"user_pref","value":"Prefers dark mode"}'

# 2. Search memories
curl -X POST https://remlabs.ai/v1/memory/search \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"query":"user preferences"}'

# 3. List memories in a namespace
curl https://remlabs.ai/v1/memory/list?namespace=default&limit=10 \
  -H "Authorization: Bearer your-api-key"
    

SDKs

Official SDKs

First-party clients for Python and Node.js. Or go direct with the REST API.

Python SDK

Full-featured Python client. Sync and async. Type hints throughout.

pip install remlabs

PyPI →

Node.js SDK

TypeScript-first. ESM and CJS. Works in Node, Deno, Bun.

npm install @remlabs/sdk (private beta)

Request beta access →

REST API

Direct HTTP. Any language, any platform. OpenAPI spec available.

https://remlabs.ai/v1/

API Docs →

Integrations

Works with your stack

One knowledge layer across every tool your agents use.

View all integrations →

MCP Server

MCP config for Claude Code & Cursor

Persistent memory in any MCP client. Paste this into .claude/mcp.json or .cursor/mcp.json.

JSON — .claude/mcp.json or .cursor/mcp.json

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

Your assistant now has memory across sessions. Store context, recall earlier conversations, build long-running workflows — all over natural language tool calls.

Dream Engine API

Background consolidation, on demand

Beyond store and search: kick off a dream cycle that runs nine consolidation strategies across your namespace and returns insights, contradictions, and new memories.

Python — Dream Engine

# Start a dream cycle — 9 strategies run autonomously
dream = rem.dream(namespace="default")
# Returns: { id: "d_abc123", status: "running" }

# Poll for completion
status = rem.dream("status", id=dream["id"])
# Returns: { status: "complete", memories_created: 7,
#   insights: ["User's auth issues cluster around token refresh"],
#   contradictions_found: 2 }

# Run a specific subset of strategies
rem.dream(namespace="default", strategies=["pattern_extract"])
# 9 strategies: synthesize, pattern_extract, insight_generate,
#   compress, associate, validate, evolve, forecast, reflect