Synap MCP server
{
"mcpServers": {
"maximem-synap": {
"type": "http",
"url": "https://synap-mcp.maximem.ai/mcp"
}
}
}{
"mcpServers": {
"maximem-synap": {
"type": "http",
"url": "https://synap-mcp.maximem.ai/mcp"
}
}
}{
"mcpServers": {
"maximem-synap": {
"type": "http",
"url": "https://synap-mcp.maximem.ai/mcp"
}
}
}{
"servers": {
"maximem-synap": {
"type": "http",
"url": "https://synap-mcp.maximem.ai/mcp"
}
}
}{
"mcpServers": {
"maximem-synap": {
"type": "http",
"url": "https://synap-mcp.maximem.ai/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 0.1.0
What it does
Synap is a memory layer for AI agents that persists conversation context and user information across sessions. It provides an MCP server and SDKs (Python and JavaScript) for agents to record and retrieve conversation history, with retrieval happening via cloud-hosted memory processing. It is for developers building multi-turn agent applications that need users' information to persist across separate conversations.
- Retrieve context about what a user was working on in a previous conversation when they return for a new session.
- Record conversation messages and user interactions to build a persistent knowledge graph across multiple conversations with the same user.
- Fetch formatted context based on search queries to inject relevant user history into an agent's current conversation.
Tools this server exposes
| Tool | Does |
|---|---|
| record_message | Record a message from a conversation with metadata like conversation ID, user ID, and message content. |
| fetch | Retrieve formatted context for a user and conversation based on search queries. |
| fetchUserContext | Get user context from across all their conversations using search queries. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.