Persistent hypergraph memory skill for OpenClaw. Give your AI agent long-term memory that persists across conversations, with rich relationship modeling between facts.
HyperMemory is a cloud-hosted knowledge graph service that stores information as interconnected nodes and relationships. Unlike simple key-value stores:
- Hypergraph Structure - Model complex multi-way relationships (not just pairs)
- Semantic Search - Find relevant memories using natural language queries
- Persistent Storage - Data stored in PostgreSQL, never lost
- Instant Recall - Sub-second retrieval across your entire knowledge graph
Option A: ClawHub (Recommended)
clawhub install hypermemoryOption B: Manual Installation
# Clone this repository
git clone https://github.com/RunStack-AI/hypermemory-openclaw.git
# Copy to your OpenClaw skills directory
cp -r hypermemory-openclaw ~/.openclaw/workspace/skills/hypermemory
# Refresh skills
openclaw skills refresh- Sign up at hypermemory.io
- Go to your dashboard and create a new API key
- Copy the key (starts with
hm_)
Via OpenClaw Secrets (Recommended)
openclaw secrets set HYPERMEMORY_URL "https://api.hypermemory.io"
openclaw secrets set HYPERMEMORY_API_KEY "hm_your_api_key_here"Via Environment Variables
export HYPERMEMORY_URL="https://api.hypermemory.io"
export HYPERMEMORY_API_KEY="hm_your_api_key_here"Via ClawCloud Console
Navigate to Settings → Environment and add:
HYPERMEMORY_URL:https://api.hypermemory.ioHYPERMEMORY_API_KEY: Your API key
openclaw agent --message "use hypermemory_overview to show my memory"The skill provides 7 tools that your OpenClaw agent can use:
| Tool | Purpose |
|---|---|
hypermemory_overview |
Get summary and statistics of your memory graph |
hypermemory_store |
Store a new memory with optional relationships |
hypermemory_recall |
Search memories using natural language |
hypermemory_update |
Update an existing memory |
hypermemory_forget |
Remove a memory |
hypermemory_find_related |
Find memories connected to a specific node |
hypermemory_add_relationships |
Create relationships between existing memories |
When properly configured, your agent will automatically:
- At conversation start: Call
hypermemory_overviewandhypermemory_recallto load relevant context - During conversation: Store new facts, preferences, and decisions as they come up
- Silently: Memory operations happen in the background without cluttering the conversation
When the user says "My name is Alex and I work at Acme Corp":
hypermemory_store({
key: "user_profile",
description: "User is Alex, works at Acme Corp",
data: { name: "Alex", company: "Acme Corp" }
})hypermemory_store({
key: "project_webapp",
description: "E-commerce webapp using Next.js and Stripe",
data: {
name: "ShopFast",
stack: ["Next.js", "Tailwind", "Stripe", "PostgreSQL"],
status: "development"
},
relationships: [
{ nodes: ["project_webapp", "user_profile"], relationship: "owned_by" }
]
})hypermemory_recall({
query: "user preferences for programming",
max_results: 10
})hypermemory_find_related({
key: "project_webapp",
max_depth: 2
})Use descriptive, hierarchical keys for organization:
| Prefix | Use For | Examples |
|---|---|---|
user_ |
Personal info | user_name, user_company, user_timezone |
preference_ |
User preferences | preference_editor, preference_language |
project_{name}_ |
Project details | project_webapp_stack, project_webapp_deadline |
decision_ |
Decisions made | decision_database, decision_hosting |
fact_ |
General facts | fact_team_size, fact_budget |
person_ |
People info | person_sarah_role, person_mike_contact |
hypermemory-openclaw/
├── SKILL.md # Agent instructions and tool documentation
├── manifest.json # Tool schema definitions (JSON-RPC)
├── index.mjs # JavaScript handlers for HTTP calls
├── openclaw.plugin.json # OpenClaw plugin configuration
├── package.json # NPM package definition
└── spec/ # Jest test suite
└── hypermemory.spec.mjs
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ OpenClaw │ │ HyperMemory │ │ PostgreSQL │
│ Agent │────▶│ Gateway │────▶│ Database │
│ │ HTTP│ (Fly.io) │ │ │
└─────────────────┘ └─────────────────┘ └─────────────────┘
│
▼
┌─────────────────┐
│ Redis │
│ (Real-time) │
└─────────────────┘
The skill makes HTTP POST requests to the HyperMemory gateway, which:
- Authenticates your API key
- Routes to your tenant's isolated data
- Stores/retrieves from PostgreSQL
- Returns results in MCP format
cd ~/.openclaw/workspace/skills/hypermemory
npm install
npm testYour credentials aren't set. Run:
openclaw secrets set HYPERMEMORY_URL "https://api.hypermemory.io"
openclaw secrets set HYPERMEMORY_API_KEY "hm_your_key"Your API key is invalid or expired. Check your key at hypermemory.io/dashboard.
You've hit your rate limit. Wait a moment or upgrade your plan.
Run openclaw skills refresh or restart the gateway:
openclaw gateway restartEnsure your API key has write permissions. Check your key scopes in the dashboard.
HyperMemory uses a simple query-based pricing model:
| Plan | Monthly Price | Queries/Month | Overage |
|---|---|---|---|
| Free | $0 | 10,000 | Hard limit |
| Developer | $29 | 50,000 | $0.50/1K |
| Pro | $99 | 500,000 | $0.50/1K |
| Enterprise | Contact Us | Unlimited | Custom |
A "query" is any interaction with HyperMemory - stores, recalls, updates, and deletes all count.
Contributions are welcome! Please:
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
- Documentation: docs.hypermemory.io
- Issues: GitHub Issues
- Email: support@hypermemory.io
This project is licensed under the MIT License - see the LICENSE file for details.
- HyperMemory MCP - MCP integration for Claude, OpenAI, n8n, CrewAI, and more
- OpenClaw - The AI assistant platform
- ClawHub - OpenClaw skill marketplace
Made with ❤️ by RunStack AI