Skip to main content

​
General

Orca Memory is a memory plugin for OpenClaw. It enables your Claw to remember context across sessions - your preferences, past decisions, code patterns, and project knowledge.
Orca Memory is an OpenClaw plugin. Once installed and configured, it:
  • Auto-Recall: Injects relevant memories before each AI turn
  • Auto-Capture: Stores important information after each turn
  • Tools: Provides manual memory management tools to your Claw
Yes. All data is encrypted in transit (TLS) and at rest. We use Convex for data storage, which provides enterprise-grade security. For maximum control, you can self-host Orca Memory.
The cloud version requires internet connectivity. For offline use, you can self-host Orca Memory on your local network.

​
Cloud

  1. Sign up at app.orcamemory.com
  2. Create a project
  3. Get your API keys (Key ID + API Key)
  4. Install the Orca Memory plugin in OpenClaw
  5. Configure with your credentials
See the Quickstart for details.
On the Free plan, requests beyond the daily limit will be rejected. Consider upgrading to Tide or Abyss for higher limits.
Yes. You can delete individual memories, entire projects, or your account. All data is permanently removed within 30 days per our retention policy.
New accounts start with a 14-day Tide trial. No credit card required.

​
Self-Hosting

  • A Convex account (free tier works)
  • Docker and Docker Compose
  • A server with at least 1GB RAM
See Requirements for full details.
The Orca Memory software is open source and free to self-host. You’ll only pay for your own infrastructure (hosting, Convex usage if beyond free tier).
Yes. The embeddings service supports sentence-transformer models. Configure it via the EMBEDDING_MODEL environment variable. Default is intfloat/e5-base-v2.
Pull the latest code and rebuild:
git pull origin main
docker-compose build
docker-compose up -d
npx convex deploy

​
Troubleshooting

Check that:
  1. The plugin is installed and enabled in OpenClaw
  2. Your API URL, Key ID, and API Key are correct
  3. Auto-Capture is enabled (or use tools manually)
  4. Your plan hasn’t exceeded its limits
Ensure:
  1. Auto-Recall is enabled in plugin settings
  2. You have existing memories stored
  3. The conversation has enough context for semantic matching
  4. Check maxRecallResults isn’t set to 0
Semantic search works best with:
  • Descriptive, natural language queries
  • Memories that have clear, detailed content
  • Appropriate tags for filtering
Try rephrasing your query or lowering the similarity threshold.
If you’re hitting rate limits:
  • Surface (Free): 500K tokens, 5K searches/month
  • Upgrade to Tide or Abyss for higher limits
  • Consider reducing profileFrequency to inject less often

​
Contact

Still have questions? Reach out: