Chat agentIntroducing chat agent

Build and deploy fully‑managed AI agents and workflows

Name: Trigger.dev
Author: Trigger.dev

Trigger.dev is the platform for building AI workflows in TypeScript. Long-running tasks with retries, queues, observability, and elastic scaling.

Start building now

16.1k | Open source

// Durable AI chat agent: survives refreshes, redeploys, and crashes
const tools = {
  searchDocs: tool({
    description: "Search the product docs",
    inputSchema: z.object({ query: z.string() }),
    execute: async ({ query }) => searchIndex(query),
  }),
  refundOrder: tool({
    description: "Refund an order",
    inputSchema: z.object({ orderId: z.string() }),
    needsApproval: true, // pauses for human approval; the run stays alive
    execute: async ({ orderId }) => payments.refund(orderId),
  }),
};

export const supportAgent = chat.agent({
  id: "support-agent",
  tools, // declared once, threaded everywhere and typed in run()
  run: async ({ messages, tools, signal }) =>
    streamText({
      // Wires the system prompt, tool approval (HITL),
      // compaction, mid-turn steering, and telemetry
      ...chat.toStreamTextOptions({ tools }),
      model: anthropic("claude-sonnet-4-5"),
      messages,
      abortSignal: signal,
      stopWhen: stepCountIs(15),
    }),
});

Build durable AI agents that survive refreshes, redeploys, and crashes. Stream to your frontend with no API routes, plus tool calling, human-in-the-loop, and full observability.

Learn about AI agents

How it works

Write tasks in your Trigger folder

Build invincible AI apps

Offload any long-running async AI tasks to our infrastructure. Create AI agents with human-in-the-loop functionality and stream responses directly to your frontend.

AI agents overview

Autonomous agent

AI agents that can use judgement to perform complex open-ended tasks.

View example

Prompt chaining

Chain AI prompts together to create multi-stage processing flows.

View example

Routing

Smart distribution of tasks to specialized AI models based on content analysis.

View example

Parallelization

Concurrent execution of multiple AI tasks for simultaneous processing and analysis.

View example

Orchestrator

Coordinate multiple AI agents to achieve complex objectives.

View example

Evaluator-optimizer

Iterative feedback system that continuously evaluates and refines AI outputs.

View example

Deploy and scale to any size

No timeouts

Write simple, reliable code and never hit a timeout.

Pay for what you use

Only pay when your code is actually executing.

No servers to manage

We deploy your tasks and handle scaling for you.

Find and fix bugs fast

Alerts for errors

Get notified via email, Slack or webhooks when your tasks or deployments fail.

Advanced filtering

Find runs fast using advanced filtering options, then apply bulk actions to multiple tasks at once.

Versioning

Each deploy is an atomic version ensuring started tasks are not affected by code changes.

Bring your tasks to the foreground

Trigger.dev Realtime

Display the status of your tasks anywhere in your app

Show the run status (in progress, completed, failed) and metadata to provide real-time, contextual information for your users as your tasks progress.

Realtime docs

Update your UI in real-time with the run status

Realtime streams

Stream LLM responses from your runs to your users

Forward streams from any provider through our Realtime API. Build AI agents with tools and context from your runs.

Realtime streams docs

Pipe streams to your frontend from your runs

True runtime freedom for developers

Unlike restricted runtimes, Trigger.dev lets you freely customize every aspect of your build process, resulting bundle, and final container image.

Python

Execute Python scripts with automatic package installation through requirements.txt

Learn more

Prisma

Copy files to the build directory, generate the Prisma client, migrate databases, and more.

Learn more

Puppeteer

Automate browser capabilities and control web pages.

Learn more

esbuild

Add custom esbuild plugins to your build process.

Learn more

FFmpeg

Add FFmpeg binaries to your project during build time, enabling video manipulation tasks.

Learn more

apt-get

Easily install any system packages you need, from libreoffice to git

Learn more

additionalPackages

Add additional packages which aren't automatically included via imports.

Learn more

audioWaveform

Produce visual renderings of audio using waveform data.

Learn more

Custom build extensions

Integrate custom tooling and project-specific requirements directly into your build pipeline.

Get started

Reliable by default

Docs

import { task } from "@trigger.dev/sdk";

export const simpleTask = task({
  id: "simple-task",
  retry: {
    maxAttempts: 3,
    minTimeoutInMs: 1000,
    maxTimeoutInMs: 5000,
    factor: 2,
    randomize: true,
  },
  run: async (payload, { ctx }) => {
    logger.log(`Attempt ${ctx.attempt.number}`);

    try {
      throw new Error("This is an error.");
    } catch (error) {
      // The error was caught, so no retry
    }

    throw new Error("This will cause a retry.");
  },
});