OpenClaw Integration Guide

What is OpenClaw?

OpenClaw is an open-source AI agent platform that works with any AI agent to empower it with memory. It connects to WhatsApp, Telegram, Discord, Slack, iMessage, and other messaging channels, with built-in local memory featuring semantic and exact match capabilities.

Neutron provides the long-term smart memory layer for OpenClaw - enabling cross-channel context retrieval, multimodal semantic search, and persistent memory that grows with every interaction.

Key Features

Auto-Recall

Before each AI interaction, Neutron retrieves relevant past conversations and user preferences to inform responses.

Auto-Capture

After each exchange, conversations are automatically stored in Neutron for long-term retention and deduplication.

Multi-Platform

Works across WhatsApp, Telegram, Discord, Slack, iMessage, Signal, Microsoft Teams, and more.

Slash Commands

Users can interact directly with memory using /remember and /recall commands.

Install from ClawHub

The Neutron Agent Memory skill is published on ClawHub — OpenClaw's public skill registry. It lets your agent store and retrieve memories using Neutron's semantic search API.

Step 1 — Install the ClawHub CLI

Terminal

npm i -g clawhub

Step 2 — Install Neutron Agent Memory

Run this from your OpenClaw project directory:

Terminal

clawhub install neutron-agent-memory

The skill will be installed into ./skills/neutron-agent-memory/. Start a new OpenClaw session to pick it up.

Step 3 — Configure environment variables

Set these in your OpenClaw .env file. The skill reads them automatically.

.env

NEUTRON_API_KEY=your_api_key_here
NEUTRON_AGENT_ID=your_agent_id_here
NEUTRON_EXTERNAL_USER_ID=1

Get your API Key and Agent ID from the Manage page.

View on ClawHub

Authentication

All API requests require:

Request Headers

Authorization: Bearer $NEUTRON_API_KEY

Tenant context is passed as query parameters: appId (your Agent ID) and externalUserId.

Seeds (Memory Storage & Search)

Seeds are the core unit of memory in Neutron. Use them to store long-term knowledge like documents, notes, conversation history, and reference material.

Save Text Content

All seed uploads use multipart/form-data with JSON-stringified arrays. The text, textTypes, textSources, and textTitles arrays must be the same length.

POST /seeds - text type

curl -X POST "https://api-neutron.vanarchain.com/seeds?appId=$NEUTRON_AGENT_ID&externalUserId=$NEUTRON_EXTERNAL_USER_ID" \
  -H "Authorization: Bearer $NEUTRON_API_KEY" \
  -F 'text=["User prefers oat milk lattes from Blue Bottle every weekday morning"]' \
  -F 'textTypes=["text"]' \
  -F 'textSources=["bot_save"]' \
  -F 'textTitles=["User coffee preference"]'

Save JSON Content

Use json type for structured data like conversation logs, user profiles, or agent state.

POST /seeds - json type

curl -X POST "https://api-neutron.vanarchain.com/seeds?appId=$NEUTRON_AGENT_ID&externalUserId=$NEUTRON_EXTERNAL_USER_ID" \
  -H "Authorization: Bearer $NEUTRON_API_KEY" \
  -F 'text=["{\"role\":\"user\",\"content\":\"Book me a table for 2 at 7pm\",\"timestamp\":\"2025-01-15T19:00:00Z\"}"]' \
  -F 'textTypes=["json"]' \
  -F 'textSources=["bot_save"]' \
  -F 'textTitles=["Restaurant booking request"]'

Seed types for agents:

text json markdown csv email claude_chat gpt_chat

Source (use bot_save for OpenClaw agents):

bot_save telegram_save upload mcp

Semantic Search

POST /seeds/query

curl -X POST "https://api-neutron.vanarchain.com/seeds/query?appId=$NEUTRON_AGENT_ID&externalUserId=$NEUTRON_EXTERNAL_USER_ID" \
  -H "Authorization: Bearer $NEUTRON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "natural language search", "limit": 10, "threshold": 0.5}'

Request body:

query (required) - Natural language search query
limit (optional) - Max results 1-100, default 30
threshold (optional) - Similarity threshold 0-1, default 0.5
seedIds (optional) - Limit search to specific seeds

Response: Array of results with seedId, content, and similarity score (0-1).

Agent Contexts (Session Persistence)

Agent Contexts let you persist session state, conversation history, and task progress between agent sessions.

Create Context

POST /agent-contexts

curl -X POST "https://api-neutron.vanarchain.com/agent-contexts?appId=$NEUTRON_AGENT_ID&externalUserId=$NEUTRON_EXTERNAL_USER_ID" \
  -H "Authorization: Bearer $NEUTRON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agentId": "my-agent",
    "memoryType": "episodic",
    "data": {"key": "value"},
    "metadata": {"tags": ["tag1"]}
  }'

Memory types:

episodic - Conversation history, decisions, actions
semantic - Domain knowledge, user preferences
procedural - System prompts, tool definitions
working - Current task state, variables

List Contexts

GET /agent-contexts

curl -X GET "https://api-neutron.vanarchain.com/agent-contexts?appId=$NEUTRON_AGENT_ID&externalUserId=$NEUTRON_EXTERNAL_USER_ID&agentId=my-agent" \
  -H "Authorization: Bearer $NEUTRON_API_KEY"

Filter by agentId, memoryType, or tags query parameters.

Usage Guidelines

Use Seeds for long-term knowledge: documents, notes, reference material
Use Agent Contexts for session state: conversation history, task progress
Search uses semantic similarity, not keyword matching - use natural language queries
Processing takes 5-30 seconds after creating seeds before they are searchable

Get Started

Ready to add persistent memory to your OpenClaw agents? Create a free account to get your API key.

Get Started Free