OpenAI
Anthropic
Gemini
LangChain
Vercel AI SDKTwo ways to go live.
Deploy agents as REST APIs your apps can call, or tools as MCP servers that AI clients discover automatically.
From serve() to production.
No YAML. No Dockerfiles. No infra to manage.
1
Define your agents and tools
A handler, a schema, whatever LLM you prefer.
2Add
Add serve()
One call wires up REST endpoints, MCP servers, validation, and docs.
3
Deploy
One command or connect GitHub. Scaling, secrets, rollbacks — handled.
index.ts
import { z } from "zod"
import OpenAI from "openai"
import { agent, tool, serve } from "@reminix/runtime"
const openai = new OpenAI()
const supportBot = agent("support-bot", {
type: "chat",
handler: async (input) => {
const response = await openai.chat
.completions.create({
model: "gpt-4o",
messages: input.messages,
})
return response.choices[0].message.content
},
})
const lookupCustomer = tool("lookup_customer", {
description: "Find customer by email",
inputSchema: z.object({ email: z.string() }),
handler: async ({ email }) => {
return await crm.find(email)
},
})
serve({
agents: [supportBot],
tools: [lookupCustomer],
})Terminal
$
The infrastructure you'd rather not build.
Everything between your code and production, handled.
REST APIs
Auth, validation, CORS, rate limiting.
MCP servers
Single endpoint, all tools discoverable.
SSE streaming
Token-by-token, with backpressure.
OAuth connections
20+ services, automatic token refresh.
Python & TypeScript SDKs
Type-safe clients for your agents.
Monitoring & logs
Tracing, errors, latency. Built in.