Inngest
Inngest is durable execution for workflows and AI — wrap functions in steps and get retries, flow control, and traces with zero infra.
If your background jobs or agents keep dying halfway through, Inngest is the lowest-friction fix on the market: wrap existing code in step.run() and retries, memoization, replay, and traces appear without a queue or worker tier to run. The 2026 releases — Sandboxes, bursty concurrency, scoped access keys, and the AI Overview cost/latency dashboard — are aimed squarely at teams shipping agents, not just cron jobs. Just watch the execution meter: every step counts, so chatty loops get expensive
Verified 12h ago · liveness 88/100 · cite: rightaichoice.com/tools/inngest
- AI agent and LLM pipeline builders who need pause/resume, evals, and per-model cost visibility
- Multi-tenant SaaS needing per-customer concurrency so one heavy user can't starve others
- Python teams replacing Celery brokers with durable background jobs on FastAPI, Flask, or Django
- Event-driven and webhook-heavy backends that need retries, replay, and full traces
- Teams that want a drag-and-drop visual workflow builder instead of a code-first SDK
- Systems requiring strict message ordering or exactly-once delivery guarantees
- Projects needing fine-grained state machine semantics — Temporal fits that better
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Inngest if you need a drag-and-drop workflow builder, guaranteed message ordering, or true exactly-once delivery semantics, or if you would rather buy a complete integration platform than add a durable-execution SDK to your codebase.
Every step is an execution, so a ten-step agent run bills ten times against your monthly quota — Free's 50k executions can disappear fast.
Free at $0/mo fits a solo developer or prototype under 50k executions with 5 concurrent steps and 24-hour traces. Pro at $99/mo, and Business at $499/mo, both billed monthly with no annual commitment, fit scaling production teams; Enterprise is custom. It sits above lightweight cron and queue tools and below Temporal Cloud, which carries a heavier infrastructure and pricing commitment.
In short
Inngest — Inngest is durable execution for workflows and AI — wrap functions in steps and get retries, flow control, and traces with zero infra. Best for AI agent and LLM pipeline builders who need pause/resume, evals, and per-model cost visibility, Multi-tenant SaaS needing per-customer concurrency so one heavy user can't starve others, Python teams replacing Celery brokers with durable background jobs on FastAPI, Flask, or Django. Free to start; paid plans from $99/mo.
No head-to-head comparison for Inngest yet — the alternatives below are the closest matches we list.
What's new in Inngest
Checked 8 days agoAcross the latest 9 updates: 4 feature updates, 2 launches, 1 community discussion and 2 news mentions.
Sandboxes open beta for all Pro plans and above
Inngest Sandboxes enter open beta for Pro and above — isolated microVMs created as steps that pause with the run rather than billing through the wait.
NEW: Bursty Concurrency — coverage for spikes
Blog post detailing bursty concurrency: when a spike exceeds your limit, Inngest runs up to 3x the ceiling for one hour a month instead of building a backlog.
Scoped access: read or write for MCP, CLI, and API keys
MCP, CLI, and API keys now support read and write permissions via OAuth or dashboard-created keys; calls without required permission return 403.
Sandboxes: isolated microVMs from any function
step.sandbox starts an isolated microVM from inside a function in under a second on the TypeScript SDK. Requires inngest 4.21.1+ and sandboxMiddleware().
Bursty Concurrency: spikes run up to 3x your limit
Pro and above accounts can exceed their concurrency limit by up to 3x for 60 minutes a month. Bursts are not charged; function-level limits override bursts.
Inngest kicks off campaign for long-running work
Inngest launched a campaign for people, apps, and agents built for the long run, framed around durable execution.
All your workflows are about to be long running. Build for it.
Architecture guidance for long-running pipelines, covering how one system can run one item or a million, for seconds or months.
Incident report: deletion cascades exhausted database connections
Inngest published a report on database connection pool exhaustion that interrupted run scheduling and event processing on September 18, 2026.
AI Overview: monitor AI calls, cost, and latency in one dashboard
AI Overview dashboard turns OpenTelemetry-derived gen_ai.* metadata into a live view of usage, token counts, estimated cost, and performance by function or model.
What people actually say about Inngest — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
6 mentions across 5 sources (Hacker News, YouTube, Stack Overflow, GitHub, Lemmy), 39 more we could not attribute · researched Oct 9, 2026.
Weighted by the 45 posts each of 5 sources contributed.
- +Notably simpler operationally than Temporal — no cluster, workers, or brokers to run
- +step.run() memoization means retries skip completed work instead of re-paying for it
- +First-class flow control: per-tenant concurrency, rate limits, throttling, batching out of the box
- +AI agent features (sessions, scorers, experiments, gen_ai.* dashboard) cover the agentic trend well
- +Works across TypeScript, Python, and Go with a coherent step-based mental model
- −Not a drag-and-drop builder — n8n-style visual users will be disappointed
- −No hard exactly-once or strict message-ordering guarantees by design
- −Most public commentary is tutorials and resumes rather than battle-tested scale stories
- −Pricing scales with steps/events; high fan-out workloads need careful cost modeling
- −Agent-kit docs and examples have rough edges (SO question about the summarizer)
- • Pricing scales with steps/events — high fan-out workflows can balloon cost unexpectedly
- • Long-running waits and agent sessions may count toward run volume
Viability Score
How well maintained and how widely used is Inngest? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Step-based durable execution via step.run() with automatic retries and exponential backoff
- Memoization skips completed steps on retry and resumes exactly where the run left off
- Pause and resume with step.waitForEvent() for minutes or months
- Per-tenant concurrency, rate limits, throttling, and batching
- Bursty concurrency: run up to 3x your limit for 60 minutes a month, bursts not charged
- Sandboxes: isolated microVMs started as steps with step.sandbox on the TypeScript SDK
- Built-in idempotency keys prevent duplicate runs from the same event
- Fan out thousands of jobs per event with parallel step execution
- Scheduled and cron jobs with missed-run recovery
- Replay past runs against new code and re-execute real events without resending them
- AI Overview dashboard for gen_ai.* usage, cost, and latency across model providers
- Attach evaluation scores to runs and steps with inngest.score() and deferred scorers
- A/B test step variants on live production traffic with group.experiment()
- Propagated Sessions group entire run trees, including child runs from agents
- Serverless workers: unlimited on every tier via HTTP endpoints, plus always-connected workers
About Inngest
Inngest is an open source durable execution SDK for orchestrating event-driven backends and AI agents. You keep your TypeScript, Python, or Go code and wrap the parts that can fail in step.run(); Inngest then executes each step durably, retries failures with backoff, memoizes completed steps, and resumes a run exactly where it stopped — whether it paused for milliseconds or months. Nothing to provision: functions run serverless, on containers, or VMs, you start locally with the Inngest CLI, and move to Cloud when production traffic arrives. The same primitives cover background jobs, durable message queues, multi-step workflow orchestration, cron and scheduled jobs, webhooks, and long-running agents. Flow control is first-class and per-tenant — concurrency, rate limits, throttling, batching, and idempotency keys — so one noisy customer cannot starve the rest. Recent additions sharpen the AI story: Sandboxes spin up isolated microVMs as steps (step.sandbox), bursty concurrency lets Pro accounts run up to 3x their ceiling for 60 minutes a month without backlog, and the AI Overview dashboard charts gen_ai.* token usage, estimated cost, and latency by function or model, with every chart linked back to the run behind it. Evaluation and experimentation live in the same system. inngest.score() attaches named scores to a run, step, or experiment variant, group.experiment() A/B tests step variants on live traffic, and Sessions group entire run trees including child runs that agents spawn. Scoped access for MCP, CLI, and API keys (read or write, 403 on insufficient permission) rounds out the operational surface. It fits engineering teams with event-driven or multi-tenant backends, Python teams replacing Celery, and anyone shipping agents that die mid-run. Look elsewhere if you want a drag-and-drop builder, strict message ordering, exactly-once delivery, or fine-grained state machine semantics — Temporal remains the closer fit there.
Behind the Verdict
Reach for Inngest when your async code already exists and you want durability without rewriting it. Function triggered by an event, several steps that can fail, some of them slow — that is the sweet spot, and the payoff is immediate: automatic retries, memoized steps so a retry doesn't re-pay for finished work, and a trace from trigger to completion. We'd pick it over rolling our own queue-plus-retry layer on day one. Multi-tenant SaaS is where the flow control earns its keep. Per-customer concurrency, rate limits, throttling, and batching mean a single heavy account can't swamp everyone else — the noisy-neighbor problem that raw queues ignore. Bursty concurrency, shipped in October 2026 for Pro and above, is a nice touch for launch days: spikes run up to 3x the ceiling for 60 minutes a month, and bursts aren't charged. AI teams get more than a job runner. Sandboxes let you run a new prompt, model, or code path in an isolated microVM on the same triggers and data as production, compare results, then ship. Scores and experiments attach to real runs rather than a side analytics stack, and AI Overview surfaces token usage, estimated cost, and latency by function or model. For agent pipelines where the model itself is the unreliable component, that combination is hard to assemble yourself. Where it bites: every step counts as an execution, so functions that call step.run() inside tight loops burn through plan limits quickly — the free tier's 50k executions goes faster than the number suggests. There are also hard boundaries. No drag-and-drop builder, no strict message ordering, no exactly-once delivery, and if you need formal state machine semantics, Temporal is the better-known tool. Air-gapped shops can't take the managed cloud's observability layer the way it's
Researching Inngest? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Inngest actually fits — and what changes day-one when you adopt it.
You wrap each LLM call and tool use in step.run(), add step.waitForEvent() for a human approval, and deploy the function to Vercel alongside your Next.js app on the Free tier.
Outcome: Failed model calls retry automatically, the run pauses for hours without holding a server, and the AI Overview dashboard shows token cost per run before you upgrade to Pro.
You migrate per-customer import jobs off a Celery worker fleet, set per-tenant concurrency and rate limits on each function, and use idempotency keys keyed on the import ID.
Outcome: One heavy customer can no longer starve the queue, duplicate webhook deliveries stop creating duplicate imports, and failed imports can be replayed without asking the customer to resend anything.
You wrap two prompt variants in group.experiment() with a weighted strategy, write an LLM-as-a-judge verdict back with inngest.score(), and defer the judge via createScorer() so it runs off the critical path.
Outcome: One variant runs per run and stays fixed across retries and replays, and cost, latency, and quality scores per variant are compared on the execution data Inngest already records.
Use Cases
- Build a durable AI agent that retries LLM calls on failure and persists state across multi-step reasoning.
- Automate lead enrichment by fanning out parallel scrapes and LLM scoring, then writing results to a CRM.
- Run a scheduled cron job that processes batches with concurrency controls to stay under API rate limits.
- Implement a user-defined workflow engine where your customers chain custom steps with retry and rollback.
- Handle inbound webhooks reliably, with automatic retries and a trace of what actually happened.
- Defer follow-up work such as scoring, notifications, or LLM-as-a-judge evals so slow calls never block responses.
- A/B test two prompt or model implementations against live production traffic and pick the cheaper or faster winner.
- Replace a Celery and broker setup with durable background jobs in FastAPI, Flask, or Django.
Models Under the Hood
as of 2026-10-16
Limitations
- Inngest is a durable execution platform, not a model provider, and it only pays off if you can express your work as steps in code.
- Each step counts as an execution, so high-step-count AI agents consume quota quickly: Free caps at 50k executions and 5 concurrent steps, Pro includes 1M executions and 100 concurrent steps, Business includes 10M and 500 concurrent steps, with add-ons available.
- Free is limited to 24 hour trace history, 500 MB span data, and 10K scores, and Datadog/advanced observability is a paid add-on on paid tiers.
- Self-hosting is possible but you lose the managed observability and turnkey features.
as of 2026-09-30
Verification history
We have re-verified Inngest 19 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 19 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Inngest tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Solo developer or prototype proving durability works before production, with a 50k execution ceiling and no credit card required.
What this tier adds
Starting tier: 5 seats, 50k executions, 5 concurrent steps, 500 MB span data, 10K scores, 24-hour trace history.
Pro
Starting at $99/mo
Ideal for
Production team scaling fast but still wanting flexibility, up to 15 seats with 1M executions and 7-day traces.
What this tier adds
Adds 20x executions (1M), 100+ concurrent steps, 1M+ queue depth, 7-day traces, and HIPAA and Datadog add-ons.
Business
Starting at $499/mo
Ideal for
Growing teams needing higher ceilings and better cost efficiency, up to 30 seats and 10M executions.
What this tier adds
Adds 10M executions, 500 concurrent steps, 25GB span data, 250K scores, and 14-day trace history over Pro.
Enterprise
Custom
Ideal for
Established organizations with security, support, and long-retention observability requirements.
What this tier adds
Adds SAML, RBAC, audit trails, HIPAA included, 90-day trace history, trace exports, and a dedicated Slack channel.
Where the pricing makes sense
The company stage and team size where Inngest's pricing actually pencils out — and where peers do it cheaper.
Free at $0/mo fits a solo developer or prototype under 50k executions with 5 concurrent steps and 24-hour traces. Pro at $99/mo, and Business at $499/mo, both billed monthly with no annual commitment, fit scaling production teams; Enterprise is custom. It sits above lightweight cron and queue tools and below Temporal Cloud, which carries a heavier infrastructure and pricing commitment.
Setup time & first value
How long it actually takes to get something useful out of Inngest — broken out by persona, not the marketing-page minute.
Solo developer: under an hour to first durable function — npm install inngest, wrap a function in step.run(), and run npx inngest-cli dev for local traces. Backend team replacing Celery: roughly a day to port the first job to FastAPI, Flask, or Django and sync the app. Multi-tenant SaaS with flow control: a few days to model per-tenant concurrency, idempotency keys, and alerting around the quotas.
Switching to or from Inngest
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Celery: move tasks into Inngest functions using the Python SDK, replacing broker config with step.run() and flow control options.
- →From cron and a job runner: convert each scheduled script into a function triggered on a cron schedule, gaining missed-run recovery and traces.
- →From a homegrown queue: keep your handlers and wrap the work in steps, then use idempotency keys where you currently dedupe by hand.
- →From AWS Lambda background workers: point your existing endpoints at Inngest functions and deploy alongside the app on Vercel, Netlify, or Cloudflare Pages.
- →From Temporal: port workflows into code-first steps, accepting per-tenant concurrency and replay in exchange for giving up declarative state machines.
- ↗To Temporal: rebuild step logic as workflows and activities if you need strict state machine semantics or a visual builder.
- ↗To BullMQ or Sidekiq: extract the step bodies back into plain queue jobs, adding your own retry and idempotency handling.
- ↗To a managed workflow product with a visual canvas: expect to rewrite orchestration logic that currently lives in your code.
- ↗To self-hosted Inngest: keep the SDK and functions but stand up your own execution and storage layer, losing the managed observability.
- ↗To provider-native queues such as SQS or Cloud Tasks: re-implement retries, memoization, and tracing yourself.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Inngest”, and we withheld 6: 6 could not be judged, because “Inngest” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Inngest.
Tools that pair well with Inngest
Common stack mates teams adopt alongside Inngest, with the specific reason each pairing earns its keep.
Mastra
Mastra is an open-source TypeScript agent framework for durable agents, typed workflows, and built-in evals.
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
OpenAI Agents SDK
OpenAI Agents SDK: a free, MIT-licensed Python framework for building multi-agent workflows with handoffs, guardrails, and tracing.
Alternatives to Inngest
View all
Mastra
Mastra is an open-source TypeScript agent framework for durable agents, typed workflows, and built-in evals.
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
OpenAI Agents SDK
OpenAI Agents SDK: a free, MIT-licensed Python framework for building multi-agent workflows with handoffs, guardrails, and tracing.
Frequently Asked Questions
Best-of guides
Used Inngest? Help shape our editorial sentiment research.