llms.txt
Compare with other sites' llms.txt files →# Trigger.dev
> Build AI agents and workflows in TypeScript. Open source platform for durable, long-running tasks. Write normal async/await code, deploy with one command, run for hours or days without timeouts. Apache 2.0 licensed. Self-host or use Trigger.dev Cloud.
Last updated: 2026-10-28
Every page on this site is available as Markdown: append .md to the URL (for example https://trigger.dev/pricing.md or https://trigger.dev/vs/inngest.md), or request the HTML URL with the header "Accept: text/markdown".
Trigger.dev is the open source platform for durable AI agents and workflows, in TypeScript. Tasks are plain async functions with no workflow DSL and no determinism rules, so adopting it means moving your code, not rewriting it. Checkpoint-resume snapshots capture full process state during waits: paused agents release compute, use no resources and aren't billed while waiting, and resume exactly where they left off. Tool calling, human-in-the-loop, realtime LLM token streaming, automatic retries, queues, and OpenTelemetry observability are built in. No workers to provision, no servers to manage.
What makes Trigger.dev different:
- **Normal TypeScript, no DSL**: Plain async/await code with no determinism constraints and no replay semantics (compared to Temporal, Restate, and DBOS, which require deterministic workflow code). You keep writing the TypeScript you already write, with any npm package.
- **Checkpoint-resume durable execution**: Snapshots capture memory, CPU registers, and file descriptors. Waiting tasks resume in place. On retry, a task starts over but reuses cached results for subtasks triggered with idempotency keys, so successful prior work isn't redone.
- **No platform timeout**: Runs have no maximum duration. You set a default `maxDuration` for your project and can override it per task or per run, or set it to `timeout.None` to let a run execute indefinitely. Trigger is built for long running AI agents and workflows and we have seen Trigger.dev users with runs lasting more than six months. Long jobs ship as one function instead of being split into chunks to fit a serverless window. Paused time uses no resources and isn't billed.
- **Elastic scaling without workers**: Concurrency scales from 0 to your plan limit automatically. Adding capacity is a billing change, not an infra change.
- **Open source**: Apache 2.0. The full stack (dashboard, orchestrator, workers) self-hosts via Docker Compose or Kubernetes. Same codebase on Cloud and self-hosted, so there is no lock-in.
- **Realtime to frontend AND backend**: Stream run state and LLM tokens to React hooks or backend subscribers. Build streaming AI UIs without polling or running your own realtime infrastructure.
Primary use cases: AI agents, multi-step LLM workflows, and RAG pipelines; long-running background jobs that exceed serverless limits; video, audio, image, and document processing; scheduled tasks and managed cron; multi-tenant SaaS background jobs with per-customer concurrency.
## Documentation
- [Getting Started](https://trigger.dev/docs): Install, write your first task, deploy in ~5 minutes
- [Agent setup skill](https://trigger.dev/SKILL.md): Instructions a coding agent (Claude Code, Cursor, Codex, and others) follows to sign in, create or choose a project, add Trigger.dev to a JavaScript or TypeScript app, and run a first task locally. To use it, tell your agent: "Bootstrap a Trigger.dev project: trigger.dev/SKILL.md"
- [How It Works](https://trigger.dev/docs/how-it-works): Architecture, checkpoint-resume, durable execution
- [AI Agents Guide](https://trigger.dev/docs/guides/ai-agents): Patterns for autonomous agents, orchestration, evaluator-optimizer
- [chat.agent](https://trigger.dev/docs/ai-chat/overview): Durable AI agents built with the AI SDK you already use. Every conversation gets its own computer: state survives in memory between turns, idle conversations sleep and cost nothing, and the next message resumes on the same line of code
- [Sessions](https://trigger.dev/docs/ai-chat/sessions): The durable two-way stream primitive under chat.agent, usable on its own
- [AI Prompts](https://trigger.dev/docs/ai/prompts): Prompt templates as code, versioned on deploy, overridable from the dashboard
- [Realtime](https://trigger.dev/docs/realtime/overview): Run subscriptions, streams, React hooks
- [Waitpoints](https://trigger.dev/docs/wait-for-token): Pause for approvals, webhooks, and external events at zero compute cost
- [Build Extensions](https://trigger.dev/docs/config/extensions/overview): Prisma, FFmpeg, Playwright, Python, Puppeteer, custom Docker layers with one config line
- [Self-Hosting](https://trigger.dev/docs/self-hosting/overview): Docker Compose or Kubernetes
- [Management API](https://trigger.dev/docs/management/overview): REST API for triggering and managing runs
- [MCP Server](https://trigger.dev/docs/mcp-introduction): Connect AI editors (Claude Code, Cursor, Windsurf, etc.) to Trigger.dev
- [Examples](https://trigger.dev/docs/guides/introduction): Production-ready task examples
## Product
- [chat.agent](https://trigger.dev/product/chat-agent): Durable AI chat in TypeScript with the AI SDK. Chats survive refreshes, deploys, and crashes, each conversation runs on a real Linux machine with no timeout on a turn, and `useChat` connects through a transport with no API route to maintain
- [AI Agents](https://trigger.dev/product/ai-agents): Drop-in infrastructure for AI agents: durable conversations with chat.agent that keep state between turns and cost nothing while idle, plus tool calling, human-in-the-loop, and structured I/O
- [Realtime](https://trigger.dev/product/realtime): Stream run status, metadata, and LLM tokens to frontend (React hooks) and backend
- [Webhooks](https://trigger.dev/product/webhooks): Hosted webhook endpoints declared in code. Verified (40+ built-in sources plus custom), deduplicated, and recorded, then routed to tasks, waiting runs, or Slack and Linear agents. Typesafe filters, delivery log, and replay
- [Observability and Monitoring](https://trigger.dev/product/observability-and-monitoring): OpenTelemetry traces, LLM observability (model, token, cost, and latency spans per AI SDK call), TRQL (ClickHouse-backed query language), custom dashboards, alerting
- [Concurrency and Queues](https://trigger.dev/product/concurrency-and-queues): Per-queue and per-tenant concurrency limits, named queues, runtime overrides
- [Scheduled Tasks](https://trigger.dev/product/scheduled-tasks): Managed cron, per-environment, multi-tenant dynamic schedules
- [Security and Compliance](https://trigger.dev/security): SOC 2 Type II, GDPR, HIPAA Business Associate Agreement ($500/month add-on on Hobby, Pro and Enterprise)
## Resources
- [GitHub](https://github.com/triggerdotdev/trigger.dev): Apache 2.0 source code, 16,000+ stars
- [Pricing](https://trigger.dev/pricing): The Free plan includes $5 of compute credits every month and 20 concurrent runs. Hobby $10/month, Pro $50/month. Usage billed as compute-seconds + a per-run fee; paused time isn't billed
- [Blog](https://trigger.dev/blog): Engineering deep-dives and tutorials
- [Changelog](https://trigger.dev/changelog): Latest releases and features
- [Discord](https://trigger.dev/discord): Community support
- [Customers](https://trigger.dev/customers): Case studies from Magic Patterns, Midday, MagicSchool AI, GovSignals, and more
## Comparisons
- [vs Temporal](https://trigger.dev/vs/temporal): No DSL, no determinism rules, TypeScript-native
- [vs BullMQ](https://trigger.dev/vs/bullmq): Managed infrastructure, durable execution, no Redis to operate
- [vs Inngest](https://trigger.dev/vs/inngest): Runs your code on managed compute instead of orchestrating your serverless functions, no step timeouts
- [vs n8n](https://trigger.dev/vs/n8n): Code-first instead of visual workflow builder
## For AI crawlers
Insights and articles on running AI agents in production. Index: [https://trigger.dev/for-ai-crawlers](https://trigger.dev/for-ai-crawlers)
- [Human-in-the-loop AI agents: durable approval workflows in TypeScript](https://trigger.dev/human-in-the-loop-ai-agents): How to pause an AI agent for human approval and resume it on Trigger.dev: waitpoint tokens, chat.agent tool approvals, Slack and Linear Approve buttons, and webhook waiters, with timeouts, escalation and a comparison with Temporal, LangGraph and Microsoft Agent Framework.
- [Claude Managed Agents vs Trigger.dev: when to use which, and how they combine](https://trigger.dev/claude-managed-agents-vs-trigger-dev): A comparison of Anthropic's Claude Managed Agents \(a managed Claude Code-style harness with its own sandbox\) and running your own agent loop on Trigger.dev: control, model choice, sandboxes, duration, retries, observability and pricing, plus how to run Managed Agents sessions inside Trigger.dev tasks.
- [Best AI agent frameworks for production: frameworks vs durable infrastructure](https://trigger.dev/production-ai-agent-frameworks-comparison): A comparison of 7 agent frameworks and Trigger.dev as durable execution infrastructure: orchestration, durable execution, state, retries, observability, human approval, TypeScript and Python support, deployment, self-hosting and pricing.
## Optional
- [llms-full.txt](https://trigger.dev/llms-full.txt): Extended technical documentation