The TypeScript framework for AI agents

Build production AI agents
in TypeScript

Redtuma gives you agents, workflows, memory, RAG, observability and MCP — the high-level primitives for real AI applications, on top of the Vercel AI SDK.

$npm create redtuma@latest

Get started Read the docs

Works with your stack

AnthropicOpenAIVercel AI SDKCloudflare WorkersDurable ObjectsHonoOpenTelemetryPostgreslibSQLMCPAnthropicOpenAIVercel AI SDKCloudflare WorkersDurable ObjectsHonoOpenTelemetryPostgreslibSQLMCP

Edge-native memory

Agent memory in a Cloudflare Durable Object — no database

Adaptive routing

Cheap model first, escalate only when needed

Agents

Tool-calling loop, structured output

Workflows

Branch, parallel, loops, suspend/resume

Memory

Working, semantic & observational

RAG

Chunk, embed, retrieve

Observability

OpenTelemetry tracing

MCP

Client & server

Everything you need, batteries included

One coherent framework instead of a dozen glued-together libraries.

Agents that reason and call tools

Give an agent instructions, a model, and tools. It runs a multi-step tool-calling loop and returns text or structured output.

import { Agent } from '@redtuma/core/agent'

const agent = new Agent({
  id: 'support',
  instructions: 'You are a helpful support agent.',
  model: 'anthropic/claude-opus-4-8',
  tools: { lookupOrder },
})

const { text } = await agent.generate('Where is order #4012?')

Built for every kind of agent

Internal agents

Automate operations, data lookups and back-office tasks with agents wired to your internal tools.

Customer-facing agents

Ship support and concierge experiences with memory, guardrails and streaming responses.

Developer platforms

Expose agents and workflows over HTTP and let your users build on top of them.

Ship your first agent today

Scaffold a project, point it at a model, and you have a streaming, tool-calling agent in minutes.