Glimind works with any MCP client — copy one config block. No signup; a free key is optional
for higher limits. Once added, your agent can call prepare_tool_call before any tool to get
the verdict, a working recipe, expected output, timeout/retry, and a live alternative in one round-trip.
claude mcp add --transport http glimind https://glimind.com/mcp
Add to your client's MCP config (e.g. claude_desktop_config.json, Cursor ~/.cursor/mcp.json, Windsurf mcp_config.json):
{
"mcpServers": {
"glimind": {
"type": "http",
"url": "https://glimind.com/mcp"
}
}
}
For clients that only speak stdio, bridge with npx -y mcp-remote https://glimind.com/mcp as the command.
code --add-mcp '{"name":"glimind","type":"http","url":"https://glimind.com/mcp"}'
npm i glimind # or: pip install glimind
import { Glimind } from "glimind";
const glimind = new Glimind(); // zero-config → https://glimind.com
const out = await glimind.wrap("server/tool",
(id) => callTool(id, args),
{ input: args, task: "search", autoRoute: true }); // health-check + failover + report
export OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=https://glimind.com/v1/traces export OTEL_EXPORTER_OTLP_PROTOCOL=http/json
See the OpenTelemetry guide and all integrations.
Get one endpoint that proxies your existing MCP servers with reliability pre-flight + failover built in — no per-call code, free, no signup:
curl -X POST https://glimind.com/v1/router/profiles -H 'content-type: application/json' \
-d '{"upstreams":[{"alias":"gh","url":"https://your-mcp/mcp"}],"failover":true}'
# → point your client's MCP url at the returned endpoint (https://glimind.com/router/mcp?profile=<token>).
Tools appear namespaced <alias>__<tool>, each with a live reliability annotation. See router docs.
curl https://glimind.com/v1/prepare/github%2Fsearch_repositories?task=search
Get a free key for higher limits: https://glimind.com/keys · full reference: docs · llms-full.txt (for IDE agents).