Last updated: 2026-09-29
Launched in 2021, Inngest is a serverless, event-driven workflow engine for background jobs, scheduled tasks, and multi-step AI pipelines that needs no queue or worker infrastructure to manage. Each function step retries independently instead of restarting the whole run, functions deploy natively alongside Vercel, Cloudflare Workers, or any Node.js backend, and the open-source core engine can be self-hosted.
About Inngest
Inngest is an event-driven durable workflow platform founded in 2021 by Tony Holdstock-Brown and Dan Farrelly. The company has raised $34M across four rounds, including a $3M seed in July 2023 from GGV Capital and a $6.1M Series A in January 2024 from Andreessen Horowitz, with a Series B in May 2025. Based in San Francisco, Inngest lets developers build reliable background jobs, scheduled workflows and AI agent tasks on serverless platforms without managing infrastructure.
The platform is serverless-first: TypeScript, Python and Go SDKs deploy alongside your existing app on Vercel, Cloudflare Workers, Hono or any Node.js backend, with the orchestration engine and dashboard hosted by Inngest rather than self-managed. The core engine is open source and can be self-hosted, though the cloud product's dashboard, managed infrastructure and observability stack stay cloud-only. Inngest added MCP (Model Context Protocol) support to its local dev server in 2026, letting Claude Code, Cursor and other AI assistants connect directly to functions for debugging. The company states in its site footer that Inngest is SOC 2 Type II compliant company-wide, with SAML, RBAC and audit trails reserved for Enterprise.
Inngest competes with Trigger.dev, which runs compute but lacks the event-driven model, Temporal, which is enterprise-grade, self-hosts natively and carries heavier infrastructure overhead, and traditional queue systems like BullMQ. Reviewers point to the combination of event-driven architecture, zero infrastructure on Vercel and a free plan as the main reason early-stage teams pick it over those alternatives. HokAI's Trigger.dev vs Inngest comparison walks through that choice step by step, and the wider field sits in the AI orchestration directory and the best AI orchestration tools roundup.
Who should not pick Inngest: teams that want visual, no-code automation between SaaS apps will get further with n8n, Zapier or Make, since every Inngest workflow is code. Teams whose core problem is retrieval or agent reasoning rather than durable execution should start with LlamaIndex or LangChain and add a durable runner later; the Inngest vs LlamaIndex guide covers where each layer ends. And teams running mission-critical, multi-language workflows at large scale will usually find Temporal's deterministic replay and native self-hosting worth its extra operational weight.
Pricing
Free: $0/month with no credit card, 5 concurrent steps, 5 seats, 3 workers, unlimited serverless workers and 24-hour trace history. When HokAI checked on 4 September 2026 this plan was labelled Hobby; on 29 September 2026 the pricing page calls it Free, and it has no metered overage, so functions pause at the quota until you upgrade.
Pro: from $99/month with 15 seats, 1,000,000 executions, 5,000,000 events, 100 concurrent steps, 20 workers and 7-day trace history. Past those allowances, executions are metered at tiered rates from $0.000050 down to $0.000015 each, an extra seat is $10, extra concurrency is $25 per 25 steps, extra events are $0.50 per million and extra evaluation scores are $1.50 per thousand.
Business: from $499/month with 30 seats, 10,000,000 executions, 50,000,000 events, 500 concurrent steps, 100 workers and 14-day trace history.
Enterprise: custom pricing, and the only plan with single sign-on through SAML, role-based access and audit logs; it also bundles HIPAA coverage, three months of traces and a dedicated Slack channel. On Pro and Business, HIPAA BAAs and the Datadog advanced-observability integration (listed at $300) are paid add-ons.
An execution is one function run or one step inside it, so a run with five step.run() calls uses six executions and the Free allowance covers roughly 8,300 such runs a month. For how those units compare with per-run billing, see the Trigger.dev and Temporal listings.
| Tier | Monthly price | What it includes |
|---|---|---|
| Free | Free | 50,000 executions a month, 500,000 events a month, 5 concurrent steps, 5 seats and 3 workers, 24-hour trace history, No credit card required |
| Pro | $99/mo | 1,000,000 executions a month, then metered, 5,000,000 events a month, 100 concurrent steps, 15 seats and 20 workers, 7-day trace history, HIPAA BAA as a paid add-on |
| Business | $499/mo | 10,000,000 executions a month, 50,000,000 events a month, 500 concurrent steps, 30 seats and 100 workers, 14-day trace history, 1-minute metrics granularity |
| Enterprise | Custom | Custom executions, events and seats, SAML, RBAC and audit trails, HIPAA included, 90-day trace history, Trace and log exports, Dedicated Slack channel |
Feature Comparison by Tier
| Feature | Free | Pro | Business | Enterprise |
|---|---|---|---|---|
| Price | $0 | $99/mo | $499/mo | Custom, quote only |
| Executions per month included | 50,000 | 1,000,000 (metered after) | 10,000,000 (add-on after) | Custom |
| Events per month included | 500,000 | 5,000,000 | 50,000,000 | Custom |
| Concurrent steps | 5 | 100+ | 500 | Custom |
| Seats | 5 | 15 (extra $10 each) | 30 (extra $10 each) | Custom |
| Workers | 3 | 20 + add-on | 100 + add-on | Custom |
| Trace history | 24 hours | 7 days | 14 days | 90 days |
| HIPAA BAA | Not offered | Paid add-on | Paid add-on | Included |
| SAML, RBAC, audit trails | No | No | No | Yes |
| SOC 2 Type II | ✓ (company-wide) | ✓ (company-wide) | ✓ (company-wide) | ✓ (company-wide) |
Key Features
- Event-Driven Workflows: Everything is triggered by events, eliminating the need to manage traditional queues or worker processes. Functions respond to events from webhooks, external services, or scheduled triggers.
- Step-Level Durability: Each step is backed by its own queue job with automatic retry on failure, saving costs and complexity compared to queue-level durability approaches used by other platforms.
- Multi-Tenant Flow Control: Built-in concurrency, throttling, rate limiting, prioritization, and debouncing controls handle multi-tenant fairness and reliability without custom implementation.
- Scheduled and Cron Jobs: Schedule jobs to run minutes or weeks into the future, enabling time-delayed workflows, deferred notifications, and reminder systems without additional infrastructure.
- Native Vercel Integration: Automatically syncs deployed code, sets environment variables, and manages infrastructure synchronization when you deploy Next.js apps, eliminating manual setup and worker processes.
- Event History and Replay: Stores full event history for observability, testing, and replay. Re-run failed workflows in bulk or test with historical events without re-triggering real upstream sources.
Pros
- The Free plan is enough for most side projects and early-stage startups to run real background jobs without paying anything upfront.
- Automatic step-level retries mean a single flaky step gets retried in isolation rather than re-running an entire multi-step workflow, which reviewers say saves real debugging time.
- MCP support lets you debug and trigger functions directly from Claude Code or Cursor, a convenience most workflow platforms don't offer yet.
Cons
- The open-source core engine can be self-hosted, but the dashboard, managed infrastructure, and full observability stack are cloud-only, so a self-hosted deployment gives up most of what makes the hosted product easy to run.
- Every piece of workflow logic has to be wrapped in explicit step.* calls with cache keys for resumability, which is more verbose than Trigger.dev's plain async function style.
- Documentation thins out fast for advanced cases like multi-region setups and integration troubleshooting, so non-standard architectures take more trial and error to get working.
- Pro pricing adds up fast once you exceed the included limits: extra seats, workers, span data, evaluation scores and concurrent steps are each billed as separate line items on top of the Pro base price.
Data Handling
- Training-data policy
- Inngest is a workflow orchestration platform, not a model provider, and does not train AI models on customer event or execution data. Trace and log history for observability is retained per plan: 24 hours on Free, 7 days on Pro, 14 days on Business, 90 days on Enterprise.
- Compliance
- SOC 2 Type II (company-wide · per Inngest's own site footer) · SAML/RBAC/audit trails on Enterprise · HIPAA BAAs as a Pro/Business add-on and included on Enterprise
Frequently Asked Questions
What does Inngest actually cost?
Inngest sells four plans: Free at $0, Pro from $99 a month, Business from $499 a month, and Enterprise at a custom price (Inngest's pricing page, checked 29 September 2026). Pro covers 1,000,000 executions and 5,000,000 events, then meters each extra execution at $0.000050 falling to $0.000015 with volume. Business covers 10,000,000 executions and 30 seats. HIPAA BAAs cost extra on Pro and Business and come included on Enterprise.
Can you use Inngest without paying?
Yes. The Free plan costs nothing, needs no credit card and does not expire, per the plan card on inngest.com/pricing; each month it includes 50,000 executions, 500,000 events, 5 concurrent steps, 5 seats and 24 hours of trace history. It has no metered overage: once the quota is used, functions pause and you need to upgrade to resume. Because every step counts as an execution, a five-step job uses six per run.
What should you use instead of Inngest?
The closest alternatives are Trigger.dev, which runs your own compute but lacks Inngest's fully event-driven trigger model, and Temporal, which offers deeper workflow control across more languages but requires running or subscribing to a dedicated cluster. Inngest's own billing FAQ notes the three bill on different units: Temporal per action, Trigger.dev per run plus compute time, Inngest per execution. Teams wanting a queue-free option on Vercel or Next.js typically pick Inngest.
Inngest or Temporal: which should you pick?
Temporal is the more capable option for complex, long-running, mission-critical workflows across many languages, with deterministic replay, and it self-hosts natively as a dedicated cluster (or via Temporal Cloud). Inngest's open-source core engine can also be self-hosted, but its managed dashboard and observability stack are cloud-only, and functions deploy directly alongside your existing Vercel, Cloudflare or Node.js app with no cluster required. Teams with deep multi-language enterprise workflows should lean toward Temporal; teams shipping typical web app background jobs in TypeScript or Python usually start with Inngest instead.
What does it take to start using Inngest?
Getting started means creating a free Inngest account, installing the TypeScript, Python or Go SDK, and writing your first function with an event trigger and one or more step calls. Running npx inngest-cli dev starts a local dev server that discovers your functions and shows traces and logs, so you can test event flows and step retries before deploying to Vercel, Cloudflare or any Node.js host. The Next.js quick start in Inngest's docs is the shortest path.
Top Alternatives
- Trigger.dev: Trigger.dev opens source with plans from $10/month; Inngest answers with zero-infrastructure Vercel integration and a free Hobby tier.
- Temporal: Temporal handles workflows-as-code with deterministic replay for resilient microservices at large scale; Inngest handles zero-infrastructure background jobs on Vercel or Node.js without a cluster to manage.
- n8n: n8n brings a free workflow automation platform built for technical teams; Inngest brings a code-first, event-driven engine built for developers shipping on Vercel or Next.js.
HokAI guides covering Inngest
- Inngest vs LlamaIndex: Why the Comparison Is the Wrong Question: Inngest and LlamaIndex both surface in AI agent searches, but they solve different layers of the stack: one runs step functions, the other builds RAG pipelines.
- Trigger.dev vs Inngest: Which Should You Use in 2026?: Trigger.dev and Inngest both push free self-hosting, but their docs warn against it in production. Real 2026 pricing and the agent-pause gap that decides this.
- OpenAI Agents API Pricing: What You Pay, What It Replaces, and When Claude or Bedrock Fits Better: OpenAI Agents API pricing has no platform fee: you pay tokens, tools and sandbox time. Costs worked out, plus when Claude Managed Agents or Bedrock fits better.
- Best AI Orchestration Tools in 2026: How to Actually Choose: n8n self-hosts free with 2,067 integrations; Zapier bills by task but lists 9,976 apps. Nine AI orchestration tools sorted into the three tiers that matter.
- AI Development Services in 2026: Which Layer You Actually Need: AI development services means four separate infrastructure layers, not one purchase: sandboxing, prompt testing, workflow durability, and API governance.