MCP Integration
Connect any agent to
Kumiho Memory.
The Kumiho MCP plugin gives AI agents persistent memory through the Model Context Protocol. Install once, remember forever.
Available MCP tools
These tools are exposed to your agent automatically when the plugin is installed.
kumiho_memory_storeStore a conversation summary with typed edges
kumiho_memory_recallRetrieve memories by semantic query
kumiho_memory_consolidateRun Dream State consolidation
kumiho_memory_ingestBuffer a message in working memory
kumiho_search_itemsSearch items by name, kind, or metadata
kumiho_create_revisionCreate a new revision on an item
Platform setup
Choose your platform and follow the setup instructions.
Claude Code
npm:kumiho-cowork- Zero-config persistent memory
- Automatic session continuity
- Memory recall across all sessions
// .claude/settings.json
{
"plugins": {
"kumiho-memory": {
"url": "npm:kumiho-cowork"
}
}
}Claude Cowork
npm:kumiho-cowork- Shared team memory
- Decision attribution to authors
- Cross-session team context
// Cowork project configuration
// Team-wide memory sharing is automatic
// Each member's decisions are attributed
kumiho_memory_recall("project architecture")
// → Team-wide decisions with full provenanceOpenClaw
pip install kumiho-memory- Full graph traversal
- Memory lifecycle control
- Agent restart persistence
from openclaw import Agent
from kumiho import KumihoMemory
agent = Agent(
memory=KumihoMemory(api_key="..."),
# Memories persist across agent restarts
# Full graph traversal for reasoning
)Ready to connect?
Pick your platform, install the plugin, and your agent gets persistent memory in minutes.