AI-native backend for real products

Build apps that think, act, and coordinate across your stack.

ArchAstro is the server side for AI-native applications: a workflow engine, an agent framework, and an integration layer exposed through Swift and TypeScript SDKs.

Start buildingRead the guide
TypeScript SDKSwift SDK

Real-time

agent updates

Resilient

workflow runs

Secure

data context

Unified

API surface

The A-ha moment

Zapier + ChatGPT + API

The infrastructure you always wished existed.

Trigger a workflow, invoke an agent, enrich with data, and ship a response — all without stitching together ten vendors.

Trigger: user signs upok
Step 1: enrich profileok
Step 2: agent drafts responseok
Step 3: human approvesok
ArchAstro SDK delivers the result

Three pillars

Everything your AI product needs to move with confidence.

Build for workflows, not single calls. ArchAstro keeps agents, data, and integrations in one composable platform.

Workflow Engine

Design multi-step automations that move from trigger to action without brittle glue code.

AI Agent Framework

Spin up task-focused agents with tools, memory, and guardrails built for real work.

Integration Layer

Securely connect external systems so agents answer with real context, not guesses.

Core concepts

The building blocks behind every AI-native app.

Workflows orchestrate multi-step logic, agents handle intelligence, integrations supply context, and client-side agents keep experiences fast and private.

Workflows

Orchestrate AI + human steps, from enrichment to follow-ups, in a single pipeline.

AI Agents

Purpose-built agents with clear goals, tools, and conversation state.

Integrations & Context

RAG-ready data connectors that keep answers grounded in your source of truth.

Client-Side Agents

Run the loop locally for privacy, speed, and deeper in-app control.

Message trajectory

See the full chain of tool calls and responses.

User: "Can you summarize Q3 revenue?"
Agent → Tool: search_finance_reports
Tool → Agent: returns top 3 docs
Agent: "Here is the summary with citations."

AI-native mindset

Quirks that make your product resilient.

Jump to SDKs

Embrace asynchronicity

Use WebSockets and stream responses so the UI feels alive while agents work.

Prompt with intent

Treat the system prompt like a constitution. Be specific, persona-driven, and iterative.

Machines calling machines

Tool design matters. Name functions clearly and build safeguards for actions.

Understand trajectory

Debug the full chain of tool calls and responses, not just the final message.

Plan for rate limits

Backoff and batch to keep the experience stable and costs predictable.

Dynamic tools

Let integrations unlock tools automatically and handle missing access gracefully.

SDKs

Multi-language SDKs that feel like first-class frameworks.

TypeScript and Swift SDKs give you the same surface area as the platform — threads, messages, workflows, integrations, and tool calls — with real-time updates baked in.

WebSockets built inStreaming responsesTooling for evals
TypeScriptarchastro/sdk
import { ArchAstroClient } from "@archastro/sdk"

const client = new ArchAstroClient({ auth: "YOUR_API_KEY" })
const thread = await client.chat.createThread({ name: "Support" })

await client.chat.sendMessage(thread.id, {
  content: "Help me with my billing issue."
})

client.chat.onMessage(thread.id, (message) => {
  console.log("Agent replied:", message.content)
})

Next steps

Go from first agent to full-stack automation.

Explore the reference apps, run evals, and ship with confidence using the same tools we use internally.

Check out the web and iOS reference apps to see the SDKs in action.
Run evals against mock providers to keep agent logic stable.
Review the SMTP proxy to understand high-scale ingestion patterns.