One Memory, Every AI Platform
Every client below runs on the same memory engine, with the same six tools underneath. Connect it locally for a private memory graph that never leaves your machine, or over the cloud to share one memory graph across every app you use. Pick yours below to see exactly how it connects — every setup here is installed for real and tested end-to-end, not just documented from official sources.
One memory graph across AI platforms
Connect any verified client over the cloud endpoint — Claude Desktop, Cursor, Doubao, and the rest — and it resolves to the same account/memory graph, no per-app silos.
Free and private, locally
The local stdio server keeps everything on your machine, encrypted at rest by default, no registration required.
The same six tools everywhere
add_memory, search_memory, forget_memory, forget_all_memory, get_graph, export_memory — identical on every client, local or cloud.
Verified clients
Every one of these has been installed for real and tested end-to-end — connecting, calling tools, and resolving to the correct memory graph. Not one is included on documentation alone.
Claude Desktop
A JSON config file for local, OAuth 2.1 for cloud — both verified end-to-end.
View setup →Cursor
The same mcp.json file — local stdio or cloud API key, one field apart.
View setup →Codex CLI
TOML config or a GUI form — for both the local server and the cloud endpoint.
View setup →
Doubao (豆包)
No config file — a point-and-click custom connector, local or cloud.
View setup →
Hermes Agent
Nous Research's open agent — one YAML file for both local and cloud.
View setup →Gemini CLI
Google's open-source terminal agent — settings.json, mcpServers.
View setup →VS Code (Copilot)
Native MCP support — mcp.json, plus a guided Command Palette flow.
View setup →Zed
Zed Industries' editor — context_servers in settings.json.
View setup →
Continue.dev
Open-source coding agent for VS Code/JetBrains — local stdio, verified.
View setup →
LibreChat
Self-hosted chat UI — librechat.yaml, both stdio and Streamable HTTP.
View setup →Dify
Hosted LLM app builder — GUI-only, remote MCP with custom headers.
View setup →Kimi CLI
Moonshot AI's terminal agent — mcp.json, plus a kimi mcp add command.
View setup →Qwen (阿里云百炼)
Alibaba Cloud's model API — code-level MCP tool, SSE, up to 10 servers.
View setup →
Cline
One of the most MCP-native coding agents — VS Code, stdio and Streamable HTTP.
View setup →JetBrains AI Assistant
IntelliJ, PyCharm, WebStorm, and more — same JSON shape as Claude Desktop.
View setup →ChatGPT
OAuth-only custom connectors — same pattern as Claude Desktop, not API key.
View setup →Amazon Q Developer CLI
AWS's coding agent — verified local and cloud, via the OAuth endpoint.
View setup →Perplexity
Hosted search AI — GUI-only custom connectors, Pro/Max/Enterprise only.
View setup →DeepSeek Harness
DeepSeek's own agent CLI — not the consumer app, still in developer preview.
View setup →LangChain / LangGraph
Code-level integration via langchain-mcp-adapters — shared by both frameworks.
View setup →
Langflow
Visual, low-code agent builder — GUI-based MCP client, no code required.
View setup →Mistral Vibe
Mistral's own coding CLI — config.toml, local and cloud both real.
View setup →Mistral Le Chat
OAuth-only custom connectors — same pattern as ChatGPT, not API key.
View setup →MiMo Code
Xiaomi's official coding agent — mimocode.jsonc, local and cloud both real.
View setup →Any other MCP client
Not one of the names above? Every MCP client speaks the same protocol underneath, so the local stdio / remote HTTP pattern below applies regardless of which one you use.