Skip to main content Overview Orca Memory is an OpenClaw plugin that provides persistent memory for your Claw. It stores and retrieves memories automatically to maintain context across sessions.
Projects A project is your workspace in Orca Memory. Each project has:
A unique API key for authentication
Its own set of memories
Configurable memory types
Usage analytics
You can create multiple projects for different codebases or use cases.
Agents An agent represents your Claw connected to a project. When you install the Orca Memory plugin, your Claw becomes capable of:
Storing new memories via the API
Retrieving relevant memories before responding
Updating existing memories as context changes
Plugin Features Auto-Recall When enabled, relevant memories are automatically injected into the conversation before each AI turn. The plugin:
Takes the current conversation context
Searches for semantically similar memories
Injects them as system context
Configure with maxRecallResults (default: 10) and profileFrequency.
Auto-Capture When enabled, important information is automatically extracted and stored after each turn. The plugin analyzes the conversation and stores:
User preferences mentioned
Decisions made
Facts learned
Patterns observed
Configure with captureMode: all (default, filters noise) or everything.
Your Claw also has manual control via tools:
Tool Description orca_memory_storeStore a new memory orca_memory_searchSearch existing memories orca_memory_forgetDelete a memory orca_memory_profileView memory profile
Memory Types Orca Memory uses four distinct memory types, inspired by human cognition:
Episodic Memory Stores specific events and conversations. Useful for:
Recalling past debugging sessions
Remembering previous discussions about architecture
Tracking decision history
Semantic Memory Stores facts and knowledge about your project. Useful for:
Project structure and conventions
Technology stack information
Business logic rules
Procedural Memory Stores patterns, preferences, and how-to knowledge. Useful for:
Coding style preferences
Workflow patterns
Repeated procedures
Working Memory Stores active, short-term context. Useful for:
Current task details
Recent decisions in the session
Temporary context
Working memory is typically cleared or updated more frequently than other types.
How It Works
User interacts with their ClawAuto-Recall fetches relevant memories from Orca MemoryClaw responds using retrieved contextAuto-Capture stores new learnings as memories 
