Skip to content
Agents, voice, images, and phone calls. Built with TypeScript.Explore the SDK
Agentium

Agentium / TypeScript AI SDK

Build your next idea.Connect it with AI.

Build agents, voice conversations, image generation, and phone calls with one TypeScript SDK. Add the tools, memory, workflows, and controls your Node.js application needs.

npm install @agentium/core

Start with core. Add the integrations your application needs.

  • One composable TypeScript SDK
  • Choose your models
  • Keep your infrastructure
  • Typed tools with Zod
  • Memory you configure
  • Teams and workflows
  • Structured output
  • Approval gates
  • Budget checks
  • Tracing and metrics
  • Evaluation package
  • MCP and A2A
  • Realtime voice
  • Image generation tools
  • Telephony adapters
  • Reusable harnesses
  • Cost accounting
  • Browser automation

Start with TypeScript

An agent.Your application.

Choose a model, give it instructions, and run it from TypeScript. Add context and tools as your application grows.

01 / 06
first-agent.ts
import { Agent, openai } from "@agentium/core";

const productGuide = new Agent({
  name: "product-guide",
  model: openai("gpt-6-astra"),
  instructions: "Explain technical ideas with one concrete example.",
});

const answer = await productGuide.run(
  "When should an agent use a tool?",
);

console.log(answer.text);

Voice, images & telephony

Give your applicationmore ways to interact.

Speak with users, create images, or manage outbound calls. Choose dedicated adapters for each capability and connect them to your application’s tools and workflows.

  • Voice
  • Images
  • Telephony
  • Build a voice conversation.

    Use OpenAI Realtime or Gemini Live, or compose speech recognition, an agent, and speech synthesis with providers such as ElevenLabs and Sarvam. Connect your audio transport separately.

    • OpenAI Realtime
    • Gemini Live
    • ElevenLabs
    • Sarvam
    • LiveKit
  • Give an agent image tools.

    Generate or edit images with OpenAI or Gemini tools. Choose the provider and supported options for your application.

    • OpenAI Images
    • Gemini Images
  • Connect outbound calls.

    Manage outbound calls through Twilio, Exotel, Telnyx, SignalWire, Vonage, or LiveKit SIP. Track call intents and status; connect voice and audio transport separately.

    • Twilio
    • Exotel
    • Telnyx
    • SignalWire
    • Vonage
    • LiveKit SIP

Build the application

The parts you need, already connected.

Build on the same foundations whether your product answers questions, creates images, handles calls, or coordinates a team of agents.

How the pieces work together

Follow one request through the stack.

Context, decisions, tools, and human review belong in the same application flow.

01/ Context

Start with the right context.

Load the conversation and retrieve the information the agent needs.

Illustrative support workflowStep 01 / 04

Incoming request

“I need help with invoice A104.”

  1. 01 Context

    Context loaded

    • Account history

      Plan
      Team
      History
      2 prior tickets

      Session memory

    • Invoice record

      Invoice
      A104
      Items
      2 line items

      Connected data

    Example data. Your storage and retrieval decide what loads.

  2. 02 RoutingNot started
  3. 03 ToolsNot started
  4. 04 ApprovalNot started

02/ Routing

Send the work to the right specialist.

Route by the request’s context, using a model or your own application logic. Give each specialist the tools it needs.

Illustrative support workflowStep 02 / 04

Incoming request

“I need help with invoice A104.”

  1. 01 ContextAccount history and invoice A104 loaded
  2. 02 Routing

    Team routing

    Router
    • Billing
    • Technical
    • General

    Use a model or application logic to choose the next specialist.

  3. 03 ToolsNot started
  4. 04 ApprovalNot started

03/ Tools

Let the agent use its tools.

Give the specialist access to the relevant API and prepare the response.

Illustrative support workflowStep 03 / 04

Incoming request

“I need help with invoice A104.”

  1. 01 ContextAccount history and invoice A104 loaded
  2. 02 RoutingBilling specialist selected
  3. 03 Tools

    Specialist at work

    lookup_invoicecompleted

    Input

    id: "A104"

    Result

    2 invoice items

    Draft response

    I found invoice A104. The second line item looks like a duplicate of the first. I can issue a credit for it once a teammate confirms.

  4. 04 ApprovalNot started

04/ Approval

Keep important actions in human hands.

Require approval before a selected tool runs, then record what happened.

Illustrative support workflowStep 04 / 04

Incoming request

“I need help with invoice A104.”

  1. 01 ContextAccount history and invoice A104 loaded
  2. 02 RoutingBilling specialist selected
  3. 03 Toolslookup_invoice ran · draft response ready
  4. 04 Approval

    Human review

    Approval gate
    issue_credit({ invoice: "A104" })Waiting for review

    The tool does not run until a reviewer decides. Nothing is called from this page.

Keep control as you grow

Know what ran. Decide what runs next.

Track usage and estimated costs across providers. Add budget checks, tool approvals, tracing, and evaluations as your application grows.

  • Approval gates

    Require a human decision before selected tools execute.

  • Costs & budgets

    Account for provider usage, estimate charges, and check configured budgets during a run.

  • Run visibility

    Add tracing, metrics, and structured logs through the observability package.

  • Evaluation

    Define test cases and score outputs before you change production behavior.

Run eventsIllustrative run
  1. run.startagent · support-desk
  2. tool.calllookup_invoice
  3. tool.resultinvoice found
  4. run.completeresponse ready
LifecycleTool callsStructured payloads

Inside a run

Follow the work, event by event.

Subscribe to agent lifecycle and tool events. Connect those signals to your logs, metrics, and traces.

Keep the run in context
Run IDs connect lifecycle events with tool activity.
Inspect what each tool did
Read tool names, arguments, and results in event payloads.
Explore lifecycle events

Fits your stack

Choose your models. Keep your infrastructure.

Choose providers for models, realtime voice, speech, images, and phone calls. Connect your existing services and storage through dedicated adapters and tools.

Explore integrations

Add the adapters and credentials you need for your chosen services.

A few things worth knowing.

Agentium is an open-source TypeScript SDK for building AI applications on Node.js. Core includes agents, tools, memory, teams, workflows, voice adapters, image tools, telephony, and cost accounting. Additional packages provide harnesses, background work, browser automation, observability, and evaluation.

No. Choose from model adapters including OpenAI, Anthropic, Google Gemini, and Ollama, or bring a custom integration. You can select separate providers for text, realtime voice, and speech. Features vary by adapter.

An Agent runs the model and tool-calling loop. The separate @agentium/harness package adds reusable context, skills, tool access, policies, budgets, and sessions around an execution driver. Use a harness when you want to share those controls across agents, teams, or workflows.

Yes, with the relevant memory configuration and storage. Configure persistence for sessions and enable additional memory features as needed. In-memory storage does not survive a process restart.

Yes. Configure approval rules for selected tool calls and connect them to your application’s review flow.

Yes. Use the dedicated voice and telephony APIs or image-generation toolkit for the capabilities your application needs. Voice supports native realtime providers and composed speech pipelines. Carrier adapters handle call control; your application connects the voice runtime and audio transport.

Enable cost accounting to inspect reported usage and estimated charges for each run. Configure budgets to check recorded costs before further requests, and add tracing and metrics through the observability package. Estimates depend on provider usage and available prices; incomplete costs stay visible.

This website describes the SDK and its runtime integrations. You choose the infrastructure on which your application runs. Use the transport and queue guides to connect it to your deployment.

Start with the quickstart, run a small agent, and then add the tools or memory your use case needs. The examples page points to focused recipes.

Start with an idea.Build it in TypeScript.

Run your first agent. Add voice, tools, and workflows as your product takes shape.