clude — Claude Code MCP server: what it does, how to install
Adds structured long-term memory with differential decay and semantic recall to AI coding agents. Useful for retaining project context across multiple sessions.
claude mcp add --env CORTEX_API_KEY=<CORTEX_API_KEY> --env CORTEX_HOST_URL=<CORTEX_HOST_URL> --env CLUDE_WALLET=<CLUDE_WALLET> --env CLUDE_AGENT_NAME=<CLUDE_AGENT_NAME> --transport stdio clude -- npx -y @clude/sdkWhat clude does
Clude supplies agents with cognitive memory instead of plain storage logs. Records belong to explicit categories — episodic, semantic, procedural, or self-model. Each category decays at an independent daily rate between 1% and 7%, while repeated retrievals strengthen the internal graph edges.
Agents connect through 8 MCP tools. These include single and batch writes up to 50 memories via batch_store_memories, hybrid search with recall_memories, and non-obvious association queries through find_clinamen. The engine can run offline using SQLite and local embeddings, or through the hosted platform with a CORTEX_API_KEY.
Direct connectors for frameworks like LangGraph or CrewAI do not exist yet. You must write custom wrappers around storage and recall methods to hook it into external orchestrators.
What it helps with
- retain architectural decisions and user context across separate CLI sessions
- retrieve relevant memories using combined vector, tag, and keyword matching
- deprecate outdated episodic details automatically via time-based decay rates
- export gathered context into JSON or Markdown files to share across models
Example prompts for the agent
- Store our decision to use Prisma migrations in procedural memory
- Recall past context regarding API rate limiting and token thresholds
- Check get_memory_stats to see the current distribution of memory types
How to install clude
Platform CLI
claude mcp add --env CORTEX_API_KEY=<CORTEX_API_KEY> --env CORTEX_HOST_URL=<CORTEX_HOST_URL> --env CLUDE_WALLET=<CLUDE_WALLET> --env CLUDE_AGENT_NAME=<CLUDE_AGENT_NAME> --transport stdio clude -- npx -y @clude/sdkReplace <…> with your own values
Platform CLI
codex mcp add clude --env CORTEX_API_KEY=<CORTEX_API_KEY> --env CORTEX_HOST_URL=<CORTEX_HOST_URL> --env CLUDE_WALLET=<CLUDE_WALLET> --env CLUDE_AGENT_NAME=<CLUDE_AGENT_NAME> -- npx -y @clude/sdkReplace <…> with your own values
Config file
{
"mcpServers": {
"clude": {
"command": "npx",
"args": [
"-y",
"@clude/sdk"
],
"env": {
"CORTEX_API_KEY": "<CORTEX_API_KEY>",
"CORTEX_HOST_URL": "<CORTEX_HOST_URL>",
"CLUDE_WALLET": "<CLUDE_WALLET>",
"CLUDE_AGENT_NAME": "<CLUDE_AGENT_NAME>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"clude": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@clude/sdk"
],
"env": {
"CORTEX_API_KEY": "<CORTEX_API_KEY>",
"CORTEX_HOST_URL": "<CORTEX_HOST_URL>",
"CLUDE_WALLET": "<CLUDE_WALLET>",
"CLUDE_AGENT_NAME": "<CLUDE_AGENT_NAME>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e CORTEX_API_KEY=<CORTEX_API_KEY> -e CORTEX_HOST_URL=<CORTEX_HOST_URL> -e CLUDE_WALLET=<CLUDE_WALLET> -e CLUDE_AGENT_NAME=<CLUDE_AGENT_NAME> clude npx -y @clude/sdkReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about clude
How do I install clude in Claude Code?
Run claude mcp add clude -e CORTEX_API_KEY=<CORTEX_API_KEY> -- npx -y @clude/sdk. Replace <CORTEX_API_KEY> with your active key.
What tools does clude expose to the model?
It provides 8 MCP tools covering single storage, batch insertions up to 50 items, hybrid recall, updates, deletions, listing, and inspection via get_memory_stats.
Can clude operate completely offline?
Yes. The architecture supports a local-first mode running on SQLite with local embeddings, eliminating external network calls and third-party API requirements.
