Integration
MCP Server
Connect Synchronex memory and decision tools to any MCP-compatible AI client.
6 min read
What It Does
The MCP server exposes memory injection, product state reads/writes, decision logging, semantic search, kernel export, session registration, and session handoff tools. It also exposes the synchronex_context prompt for one-click session context injection.
Client Configs
Claude Code should use project .mcp.json as the primary target. Cursor, Windsurf, and generic MCP clients can use the same HTTP server configuration shape.
{
"mcpServers": {
"synchronex": {
"type": "http",
"url": "https://synchronex.ai/api/mcp",
"headers": { "Authorization": "Bearer sx_live_YOUR_KEY" }
}
}
}| Client | Where to put it |
|---|---|
| Claude Code | Project .mcp.json |
| Cursor | MCP server settings |
| Windsurf | MCP server settings |
| Generic MCP client | Any config that supports HTTP MCP servers with headers |
Context Prompt
Invoke synchronex_context at the start of a session. Optionally pass product_name to append product state and handoff context.
{
"name": "synchronex_context",
"arguments": { "product_name": "Sendinel" }
}Rate Limits
MCP calls use the same tier-based limits as API calls. Free starts at 500/day, Pro at 10,000/day, and managed tiers can support higher worker-driven usage.
See the full MCP tool reference for every input schema.