Durable-execution platform for AI workflows and agents — write step-functions in TypeScript or Python, get retries, scheduling and observability for free.
Durable-execution platform for AI workflows and agents — write step-functions in TypeScript or Python, get retries, scheduling and observability for free.
Inngest is the durable execution engine that AI teams pick when they realise their long-running agents and RAG pipelines outgrow a single request-response cycle. You write workflows as ordinary TypeScript or Python functions, mark every external call as a step.run, and Inngest takes care of the rest: retries with backoff, checkpointing so failures resume where they left off, scheduled triggers, concurrency limits, parallel fan-out/fan-in, debounce, throttle, and step.waitForEvent for human-in-the-loop flows that pause for hours or days. For AI specifically, Inngest now ships AgentKit, a TypeScript framework for building multi-step agents (tool use, planning, memory) that runs on the same durable substrate, so an agent crash mid-loop doesn't drop the user's request. Inngest works with Next.js, Vercel, Cloudflare Workers, AWS Lambda and any container; the platform handles queues, scheduling and a beautiful dashboard with traces, logs and replay. Pricing is generous: free tier with 50k runs/month, Basic at \$30/month, Pro at \$150/month, with enterprise tiers for SOC 2 and dedicated infra. Inngest has become the default 'background job + agent runtime' for serverless-first AI startups.
Was this helpful?
Wrap any function call in `step.run('name', async () => ...)` and Inngest treats it as a code-level transaction with automatic retries on failure and exactly-once execution on success. State persists between steps, so multi-step workflows resume from the last successful checkpoint rather than restarting. This eliminates the need for custom idempotency keys, manual state tables, or dead-letter queues.
AgentKit is Inngest's dedicated SDK for AI agents, offering durable multi-step LLM orchestration with built-in retry on model failures and rate-limit handling. The `step.ai` primitive automatically traces every prompt/response pair into the Inngest dashboard, giving you full visibility into agent decisions without separate observability tooling. It works with OpenAI, Anthropic, and any LLM provider via standard SDKs.
Native primitives let you cap concurrency with dynamic keys (e.g., max 5 concurrent runs per user_id), throttle to respect external API rate limits, batch events for efficient bulk processing, and prioritize VIP traffic over background jobs. Built-in fairness controls eliminate noisy-neighbor issues so one heavy user can't starve others. All controls are declarative in function config rather than requiring custom queue infrastructure.
A single command (`inngest dev`) starts a local server with the same durability guarantees, retry semantics, and observability as Inngest Cloud. You can trigger events, inspect step-by-step execution, and replay runs locally before deploying. This eliminates the common production-only bug class where workflows behave differently in cloud than on a developer's machine.
When a bug ships or an upstream service fails, Replay lets you re-run thousands of failed workflows in bulk after the fix is deployed — no need to maintain dead-letter queues or write custom replay scripts. Bulk Cancellation lets you halt runaway workflows across thousands of runs with a single action. Combined with Inngest Cloud's alerting and metrics, these tools dramatically shorten incident response time.
$0
$30/month
$150/month
Contact sales
Ready to get started with Inngest?
View Pricing Options →We believe in transparent reviews. Here's what Inngest doesn't handle well:
Weekly insights on the latest AI tools, features, and trends delivered to your inbox.
Inngest is promoting its '2026 AI in Production Benchmark Report' on the homepage, surveying what the most confident teams are using to build AI. AgentKit continues to expand as a first-class SDK alongside the language SDKs (TypeScript, Python, Go, Kotlin), and step.ai prompt/response tracing is now a core observability primitive for AI workflows.
Enterprise Agents
Enterprise durable execution platform designed for AI agent orchestration with guaranteed reliability, state management, and human-in-the-loop workflows.
AI workflow infrastructure
an open-source TypeScript platform for building and deploying long-running AI agents and workflows with retries, queues, observability, realtime updates, and elastic scaling.
Automation & Workflows
Python-native workflow orchestration platform for building, scheduling, and monitoring AI agent pipelines with automatic retries and observability.
No reviews yet. Be the first to share your experience!
Get started with Inngest and see if it's the right fit for your needs.
Get Started →Take our 60-second quiz to get personalized tool recommendations
Find Your Perfect AI Stack →Explore 20 ready-to-deploy AI agent templates for sales, support, dev, research, and operations.
Browse Agent Templates →AI agents without memory restart from zero every conversation, wasting time and money. Here's how the three types of agent memory work, why they matter for your business, and which tools actually deliver results in 2026.
Deploy AI agents to production with confidence. Covers containerization, cloud deployment on AWS/Azure/GCP, Kubernetes orchestration, observability, cost control, and security best practices.
Running an online store means juggling product listings, customer questions, inventory, pricing, and marketing — all at once. AI agents can now handle most of it for you. Here's exactly how to automate your e-commerce business without hiring a team.
Compare GPT-4o, Claude 3.5 Sonnet, Gemini 2.0, Llama 4, and more for AI agent workloads. Covers tool calling, reasoning, cost, latency, and which model fits your use case.