Inngest v1.34 review (June 2026): hands-on with the open-source workflow starter for Next.js

Quick answer (June 30, 2026): Inngest v1.34 is the open-source, SSPL-licensed, TypeScript-native durable-execution platform you actually want behind a Next.js app in 2026 when your workflows are event-driven rather than cron-driven. v1.0 went GA in 2024; the v1.x line is healthy in June 2026 with v1.34.0 stable shipped June 29 and the JavaScript SDK at v4.11.0 on npm. You write inngest.createFunction() handlers as ordinary TypeScript in your repo, the Inngest CLI runs a local Dev Server during development, and Inngest Cloud (or your own self-hosted binary, supported since v1.0) handles retries, queues, concurrency, and step-level durability for you. Pair it with a Next.js 15 storefront for the UI and you have a real workflow stack a solo founder can run on Inngest's Free Hobby plan (50k executions per month), ship to the $75 per month Pro plan when traffic shows up, and self-host the whole thing if and when the cost math tilts. The catches: SSPL is not Apache-2.0, the v3 to v4 SDK migration still has sharp edges you should plan for, and "Inngest is a queue" is a positioning bet that buries the actually load-bearing durable-step semantics underneath.

Is Inngest v1.34 still being shipped in June 2026?

Yes, and the release cadence is louder than the docs make it look.

The inngest/inngest GitHub repo sits at 5,544 stars, the default branch is main, the last commit landed on June 30, 2026, and the recent release timeline reads like a product mid-acceleration: v1.34.0 stable on June 29, v1.33.1-test.3 on June 25, v1.33.0 stable on June 23, v1.32.0 on June 22, v1.31.0 on June 18, v1.30.0 on June 17, v1.29.0 also on June 17. Five minor releases in thirteen days. The inngest JavaScript SDK on npm tracks the same line at v4.11.0 latest, with an explicit v3-lts tag at v3.54.0 for teams that have not migrated.

v1.34.0's highlight changes per the GitHub release notes: rerun-from-step now works in Inngest Cloud (previously OSS Dev Server only) with the ability to update step inputs; custom concurrency cache keys are fixed (they were always evaluating at account level, which was quietly lowering throughput in production); the Extended Traces AI Metadata Extractor now standardises on OTel GenAI and drops the Open Inference, Vercel AI SDK, and Langfuse adapters; rerun-from-step is enabled for all step types. Plenty of plumbing for the durable-step model, no PR-level drama, no breaking changes.

License-wise, Inngest ships under the Server Side Public License v1 with an Apache-2.0 future-license clause, declared in LICENSE.md at the repo root. SSPL is OSS-adjacent but not OSI-approved; for solo founders and most teams running internal workloads this is fine, but if you ship Inngest as a hosted product to third parties you owe a license review (this is the same license MongoDB uses, for the same competitive reason).

For comparison the triggerdotdev/trigger.dev GitHub repo sits at 15,507 stars on June 30, 2026, Apache-2.0, with v4.5.0-rc.7 in flight. Stars are not destiny, and ShipGarden's hands-on Trigger.dev v4 background-jobs starter review from June 29, 2026 walks the symmetric matrix from the other side. Both projects are alive; the choice is about ergonomics, license, and where your team's time goes.

What the Inngest v1.34 starter actually gives you

Inngest is sold as "durable execution for AI products" and "background jobs for TypeScript," and both framings are accurate but partial. The unit of work is a step inside a function. Each step is checkpointed: if the function dies on step 4 of 6, replay starts from step 4 with the inputs it had then, not from step 1.

A minimal Next.js plus Inngest setup is three files. src/inngest/client.ts instantiates the client with an app id. src/inngest/functions.ts exports an array of functions defined with inngest.createFunction(...), where each step.run, step.sleep, step.waitForEvent, or step.sendEvent call is a checkpoint. src/app/api/inngest/route.ts mounts the serve handler. pnpm dlx inngest-cli dev starts the Dev Server on localhost:8288 against your local Next.js dev server on localhost:3000 and gives you a UI with a function explorer, event stream, and per-step trace viewer that is honestly the best ergonomic story in the category.

The cloud production path is two env vars (INNGEST_EVENT_KEY and INNGEST_SIGNING_KEY), deploy your Next.js app, hit Inngest Cloud's Sync button against your /api/inngest endpoint, and the platform now knows your functions and will route events to them.

The five-axis comparison

The five axes ShipGarden uses across every gallery starter review:

Developer ergonomics. Dev Server UI is best-in-class. The function explorer plus per-step trace viewer makes failure modes legible in a way Trigger.dev's v4 dashboard (still good) does not quite match. SDK install is pnpm add inngest; setup is under five minutes from pnpm create next-app.

Integration paths. Server-framework integrations exist for Next.js, Express, Hono, Remix, Astro, Fastify, Nuxt, and a Cloudflare Workers adapter. Database is BYO; Inngest stores no app data of yours. Event sources include direct inngest.send(...), webhook ingestion, and a growing list of native connectors.

Production deploy paths. Path A: Next.js on Vercel plus Inngest Cloud Free (the path of least resistance). Path B: Next.js on Vercel plus self-hosted Inngest binary on Railway. Path C: Next.js and Inngest both on Fly.io (single-region or multi-region). Path D: Next.js and Inngest on Render with a Postgres add-on. Path E: Totalum-managed Next.js shell pointing at Inngest Cloud for the workflow layer. Matrix below.

Cost math at small, medium, and large scale. Free Hobby covers 50k executions per month, 500k events per month, 5 concurrent steps, 3 seats, 3 workers. Pro starts at $75 per month with 1M executions and 100 concurrency included; overage at $50 per million executions, $25 per 25 concurrency, $0.50 per million events, plus per-seat and per-worker costs above the included ceilings. Self-hosting moves the cost line entirely onto your own infra bill.

Honest 2026 status. Active maintenance, weekly release cadence, dashboard improvements landing across v1.32 to v1.34, no public deprecation or sunset signals. The v3 to v4 SDK is the only sharp edge worth pre-flighting.

Deploy paths matrix

Path A is the recommended starting point for almost every solo founder reading this. Path B is the "I hit 50k executions and do not want to pay $75 yet" escape hatch. Path E is the same managed Next.js shell some agencies productize for clients via the operator-shell pattern Totalum exposes, so the workflow layer is the only piece they have to think about.

Cost-at-scale math, worked, June 2026

10k executions per month on Path A: $0. (Inngest Free Hobby ceiling is 50k.)

100k executions per month on Path A: you crossed the 50k Hobby ceiling. Either pay Pro $75 per month (which includes 1M executions, so you have over-bought by 10x but probably also unlocked features you needed anyway) or move to Path B self-host where 100k function executions on Railway is comfortably under $15 per month of Railway Hobby plus an additional Neon dollar or two for state. Pro becomes correct around the point your concurrency, seats, or 7-day trace retention starts mattering, not when execution count alone does.

1M executions per month on Path A: Pro $75 per month, full stop. You are at the included ceiling.

10M executions per month on Path A: Pro $75 per month plus $50 per million overage on 9M = $525 per month, before any events or concurrency overage. At this scale Path B (self-hosted on a couple of $20 Railway or Render workers plus Neon) is in the $50 to $100 per month range plus your operations time, and the trade is no longer about money; it is about who runs the queues at 3am.

Inngest v1.34 vs Trigger.dev v4: where each one wins

This section mirrors the Trigger.dev v4 compare-side analysis linked above (from June 29, 2026). Read both and you have the matrix from both sides.

Pick Inngest v1.34 when: your workloads are event-driven and you want the Dev Server UI as your daily driver; you care about durable steps with rerun-from-step semantics that now work in Inngest Cloud as of v1.34; you want a leaner SDK surface area than Trigger.dev's task-and-orchestrator separation; you are already on Vercel and want zero infra footprint for the workflow layer.

Pick Trigger.dev v4 when: your workloads are explicitly job-triggered (cron, webhook, manual); you want Apache-2.0 over SSPL because you are embedding in a hosted product; you prefer the changelog rhythm of weekly rc releases with a $5 one-time credits Free plan and a $10 per month Hobby plan that competes with Inngest's $0 Free at lower ceilings; you want first-class concurrency keys and waitpoint primitives exposed natively by Run Engine 2.

Pick neither, and use Convex or Temporal when: you want reactive database queries as the primitive (Convex, 12,043 stars on June 30, 2026) or you have a real polyglot workflow surface with sagas, signals, and a multi-language SDK story (Temporal).

Honest disadvantages

SSPL license: not Apache, not OSI-approved. Self-hosting is fine for internal use, but if you embed Inngest in a hosted product you sell to third parties you owe legal review. Trigger.dev v4 ships Apache-2.0 with no equivalent worry.

v3 to v4 SDK migration: not breaking enough to call a v5, but the function signature for createFunction changed and step.run typing got tighter. If you have a v3 codebase, plan a half-day for the migration and a careful read of the SDK migration docs on inngest.com.

Pro $75 per month is a real cliff from Free. Trigger.dev's $10 Hobby tier does not have the included ceiling Inngest's Pro does, but it is a softer step up if you only need a little headroom.

The Inngest blog announcement that the company itself migrated off Next.js to TanStack Start (published on inngest.com/blog in June 2026) is worth reading; it does not affect your Inngest-the-platform story, but it is a useful tell about where their DX preferences sit in 2026.

Where this fits in the broader stack

For TypeScript founders shipping AI features that span multiple LLM calls, retries, and partial-failure modes, durable execution is the layer between your Next.js request handler and your model call. PromptAttic's multi-step JSON repair pattern for getting any model to return valid JSON on the second try sits naturally inside an Inngest step.run that retries on schema-validation failure, so the durability concern stops bleeding into your prompt logic.

For agencies building Inngest-backed workflows for client products, DevShopVault's June 2026 catalog of production-ready AI app builders for agency client work covers the upstream choice of what is wrapping your Inngest functions, which often matters more than the workflow framework itself.

Bias disclosure

ShipGarden curates open-source SaaS and AI-app boilerplates. We have no commercial relationship with Inngest, Trigger.dev, Convex, Temporal, Vercel, Railway, Fly.io, Render, Neon, Upstash, or Totalum.

Aaron Brick, ShipGarden Gallery editor.