n8n pricing

About

n8n is a workflow automation platform aimed at technical teams — developers, IT, and ops engineers who want to build and run integrations and AI agents with code-level control rather than the click-only flows of Zapier or Make. Workflows are assembled on a visual node canvas, but every step can drop into JavaScript or Python, call arbitrary HTTP/GraphQL endpoints, run bash scripts, or import cURL commands, so the product sits between no-code and full custom code. The company markets itself as “the world’s most popular workflow automation platform for technical teams” and claims usage across 34% of Fortune 500 companies.

Founded in 2019 by Jan Oberhauser and headquartered in Berlin, n8n spent its first years as an open-source “Zapier alternative” before the 2023–2025 generative-AI wave turned it into an AI-agent and orchestration layer. That repositioning was rewarded in October 2025 with a $180M Series C at a $2.5B valuation (led by Accel, with NVIDIA’s NVentures, Meritech, and Redpoint participating; ~$240M raised to date) on reported 6× user and 10× revenue growth over the prior year. Its closest comparables are Zapier and Make on the automation side and a growing field of agent frameworks on the AI side.

The product ships in two forms: a managed n8n Cloud (hosted by n8n) and a free, open Community Edition that teams self-host under a fair-code Sustainable Use license. The same codebase powers both, so n8n competes against its own free self-hosted tier on managed hosting, governance, and support rather than on core features. Enterprise deployments lean on self-hosting plus paid governance: SSO/SAML/LDAP, external secret-store integration, environments with Git version control, log streaming, and SOC 2 audited infrastructure.

The defining commercial choice is the value metric. n8n charges by monthly workflow executions — one completed run of a workflow, start to finish, regardless of how many nodes or API calls it triggers — and bundles unlimited users and unlimited workflows into every paid plan. That puts the entire price gradient on a single usage dimension and deliberately avoids the per-seat and per-step meters that competitors use.

Pricing summary : How n8n’s execution-tiered workflow-automation pricing works

n8n uses an execution-tiered subscription built on a single usage dimension:

1. Flat monthly subscription per tier: Cloud plans are Starter, Pro, and Business, billed monthly or annually (annual saves 17%). Billed annually the monthly-equivalent prices are $20 / $50 / $800; billed monthly they are $24 / $60 / $960. Enterprise is custom-quoted.
2. Workflow executions as the value metric: each tier includes a fixed bucket of monthly workflow executions — 2,500 (Starter), 10,000 (Pro), 40,000 (Business) — where one execution is a full workflow run regardless of step count. There is no per-step, per-task, or per-user charge; users, workflows, and integrations are unlimited on every paid plan. The Pro and Business cards expose an execution-volume selector, so the headline bucket is the entry point of a larger ladder rather than a hard ceiling — Pro’s second step is 50,000 executions at $120/mo billed annually ($144/mo billed monthly), which is more executions than the Business bucket at a fraction of the Business price.
3. A second, non-purchasable meter — AI Assistant credits: every cloud plan now carries a monthly AI-credit allowance — 2,300 on Starter and “up to 13,700” on Pro (n8n’s FAQ says “5,700 or 13,700 on Pro, depending on your plan size”; free trials get 800 credits that expire with the trial). Credits refresh monthly, do not roll over, and cannot be bought separately — “upgrading to a higher plan is the way to get a larger allowance.” Business and Enterprise read “AI Assistant coming soon”; on self-hosted n8n you bring your own API key, “so no credit limits apply.”
4. Overage above quota is contracted, not self-serve: exceeding a Business/Enterprise quota does not stop workflows — n8n’s sales team reaches out, and unaddressed overages are “automatically invoiced 45 days after they have occurred,” with Business overage priced at 4,000 EUR per extra bucket of 300,000 executions.
5. Free and discounted entry points: a free cloud trial (no card on Starter/Pro; card required for the 14-day Business trial), a free self-hosted Community Edition, and a Startup plan at $333/mo billed annually (n8n’s “50% off Business”, 40k executions) for companies under 20 employees and under €5M funding. Note that unlike the four cloud cards, the Startup headline is an annual-cadence figure — the plan’s own Monthly/Annually toggle prices the monthly commitment higher — and n8n’s “50% off” holds against its EUR Business price, not the $800/mo USD one.

What makes this different: n8n runs a tier-bucketed usage model layered on a freemium self-host base — the price gradient sits entirely on completed workflow runs rather than per step or per seat, so a workflow that fans out across hundreds of nodes still counts as one billable execution, and headcount never moves the bill. The one exception is the new AI-credit allowance, which is deliberately un-buyable: it is an upgrade lever, not a revenue meter.

Pricing by product

n8n Cloud (subscription plans)

Self-hosted & discounted plans

Sales motions across products: PLG / self-serve for Starter, Pro, Business, the Startup plan, and free Community Edition; sales-led for Enterprise.

The on-card execution-volume selector

The four headline cards are not the whole price list. Pro and Business render their included-executions box as a dropdown, so a buyer steps the quota up without leaving the pricing page — and the price steps with it. Pro’s ladder has two rungs:

That second rung is why the Pro card advertises “Up to 13,700 AI credits/month” rather than a flat number, and why the card’s concurrency tooltip reads: “You can run 20 executions concurrently at the same time with 10k available executions. This increases to 50 concurrent executions at the same time for 50k available executions.” Note that Pro at 50,000 executions carries more monthly executions than the Business bucket (40,000) at a fraction of the Business price — Business is bought for SSO/SAML, Git version control, environments and self-hosting, not for raw execution volume. Business runs its own, much longer selector ladder (40k → 500k) whose higher rungs are quoted per-currency at checkout and are not published as static figures.

AI Assistant credits (the second meter)

Alongside executions, every cloud plan now ships a monthly AI Assistant credit allowance, replacing the smaller “AI Workflow Builder credits” line that plan cards carried through mid-2026. n8n’s own FAQ states the mechanics verbatim: “Each plan comes with a monthly allowance of AI credits: 2,300 on Starter and 5,700 or 13,700 on Pro, depending on your plan size. Free trials include 800 credits, which expire when the trial ends.” Credits are consumed by AI Assistant usage, the balance “refreshes every month, and unused credits don’t roll over,” and — the load-bearing detail — “It’s not yet possible to buy additional credits, so if you regularly need more, upgrading to a higher plan is the way to get a larger allowance.” The plan cards for Business and Enterprise read “AI Assistant coming soon”, and the comparison table lists AI Assistant credits as N/A for Enterprise. On self-hosted n8n, AI Assistant is “coming soon” with bring-your-own API key, “so no credit limits apply.”

What happens at the quota

Executions above a Business or Enterprise bucket do not stop the platform: “your workflows will continue running without interruption, but overage charges may apply if you do not upgrade to the next usage tier.” n8n’s sales team reaches out as a customer approaches the limit, and “if there’s no response or upgrade path, the overages if any will be automatically invoiced 45 days after they have occurred.” The one published overage rate is for Business: 4,000 EUR for extra buckets of 300,000 executions. A license key can be applied to an unlimited number of instances, but “the combined usage from all instances where the license key is applied will count toward your quota.”

The value metric: workflow executions, not steps

A workflow execution is one run of a workflow from start to finish — n8n counts it once no matter how many nodes, API calls, or branches it triggers. This is the structural opposite of per-task (Zapier) or per-operation (Make) metering: a 200-node workflow and a 2-node workflow each consume exactly one execution. Quotas reset monthly, and the included buckets are 2.5k → 10k → 40k across Starter → Pro → Business, with custom volume on Enterprise. Concurrency (how many executions run in parallel) is a separate per-tier limit: 5 → 20 → 200+.

Hidden costs : What execution quotas really cost at scale

The advertised $50 Pro headline understates what a growing automation team pays once execution volume climbs, because n8n’s quotas are fixed buckets rather than smooth overage. Two patterns drive most of the surprise: outgrowing the Pro execution bucket, and the gap between “free” self-hosting and its real operating cost. The first one has a cheap answer that the pricing page does not show you by default.

Archetype 1 — a team that outgrows the 10k Pro quota. A small ops team runs an AI-agent workflow that polls inboxes and triggers a model on each new message. At a few hundred messages a day they comfortably fit Pro’s 10,000 monthly executions. Add a second polling trigger and a Slack bot and they cross 10k mid-month. Reading the four headline cards, the next step looks like Business at $800. Reading the Pro card’s execution-volume dropdown, it is the same Pro plan at 50,000 executions:

So the expensive path is real but avoidable: Business costs 6.7× more than Pro-50k while including 10,000 fewer executions, because Business is priced for SSO/SAML, Git version control, environments, and self-hosting — not for volume. The genuine hidden cost here is discovery. The $120 rung renders only after a buyer opens a dropdown they have no reason to open, and the comparison table lists default buckets only, so the Reddit complaint that followed the 2025 repricing — polling triggers and chatbots “burn through execution limits in days,” pushing users onto a tier whose price rivals enterprise software — describes a cliff most of those teams could have stepped around for $70/mo.

Archetype 2 — the “free” self-hosted Community Edition. Self-hosting removes n8n’s execution meter entirely, but moves the cost onto your own infrastructure and engineering time. A modest production deployment for a single team:

So “free” self-hosting often lands near the cost of a paid cloud tier once you price the operational burden — which is exactly the value gap n8n sells managed Cloud and the self-hosted Business license into. For high-volume automation, self-hosting still wins on raw cost-per-execution; for low-to-moderate volume, managed Cloud is usually cheaper than the loaded cost of running it yourself.

Want to estimate your own n8n bill? Use the n8n pricing calculator to model your monthly cost based on workflow-execution volume and plan tier.

Pricing evolution : From open-source roots to execution-tiered cloud

n8n’s pricing started as a free, self-host-first open-source product, then layered a managed cloud on top, then in 2025 pivoted the whole model onto a single execution meter and repositioned around AI agents. In July 2026 a second dimension reappeared on the plan cards — an AI Assistant credit allowance — but pointedly not a second billing dimension: the allowance cannot be purchased at any price, so it differentiates tiers without ever generating a line item. The dollar amounts on the cloud cards did not render in the Wayback archive across any snapshot (n8n loads the price numeral via client-side localization), so the historical figures are marked unknown; what the archive does preserve clearly is the tier structure, execution quotas, and active-workflow caps.

Cadence

Tracked range: 2022-03 – 2026-07. “unknown” price-change cells mark quarters where the archive preserved structural/quota changes but the cloud dollar figures never rendered in the Wayback snapshot — they were not independently legible and are not guessed.

Notable changes

• 2022-03-17 — License changed from Apache 2.0 + Commons Clause to the Sustainable Use License (based on Elastic License 2.0); n8n coins the term “fair-code.” (n8n blog.)
• 2022 Q2 — Wayback shows the pricing page led with two free self-host options (Desktop, Self-Hosted) and a separate Cloud early-access tab.
• 2023 Q3 — Cloud paid tiers (Starter/Pro/Enterprise) carried execution quotas (5k/10k) and active-workflow caps (5/10); Starter was single-user.
• 2024 Q1 — Starter’s execution bucket dropped to 2.5k; annual toggle advertised 20% off.
• 2025-08-07 — n8n removed active-workflow limits, made users/workflows/steps unlimited on every plan, added the Business tier, and put the entire price gradient on monthly executions. (n8n blog.)
• 2025-10-09 — $180M Series C at a $2.5B valuation (Accel, NVIDIA NVentures); n8n reframes from “Zapier alternative” to AI-agent orchestration. 235-point Hacker News thread.
• 2026-06-02 — Execution-tiered Starter/Pro/Business at 2.5k/10k/40k; annual saving now 17% (down from 20%); Startup plan $333 (50% off Business, self-hosted).
• 2026-07-22 — Packaging change with no price move: the bundled AI entitlement becomes AI Assistant credits (preview) and jumps from 50 to 2,300/month on Starter and from 150 to “up to 13,700”/month on Pro (n8n’s FAQ: “5,700 or 13,700 on Pro, depending on your plan size”). Enterprise’s old 1,000 cloud-only credit line is gone — Business and Enterprise read “AI Assistant coming soon”. Credits refresh monthly, do not roll over, and cannot be purchased. The same capture surfaces two mechanics the page had never carried: a published Business overage of 4,000 EUR per extra 300,000-execution bucket, auto-invoiced 45 days after it occurs, and an execution-volume selector on the Pro and Business cards.

The 2025 repricing in detail

The August 2025 change is the most consequential moment in n8n’s pricing history and the cleanest example of the platform committing to a single value metric. Before it, every cloud tier bundled three constraints at once: an execution quota, an active-workflow cap (5 on Starter, 10 on Pro), and — on Starter — a single-user limit. n8n’s own framing was that workflow-based caps “discouraged enterprise builders from creating and experimenting,” because every new automation counted against a hard ceiling. The fix was to delete two of the three meters: workflows, users, and steps all became unlimited, leaving only monthly executions to scale the bill. The same release introduced the Business plan to fill the long gap between Pro and Enterprise, initially as a self-hosted license. Community reaction was mixed — builders welcomed unlimited workflows, but a recurring Reddit complaint is that polling triggers and chatbots can exhaust an execution bucket within days, turning a complexity-neutral metric into a volume-sensitive one for always-on AI agents.

What’s unique : Execution-metered, seat-free workflow automation

1. Priced per completed workflow execution, not per step or seat. One execution counts once regardless of node count, so complex multi-node workflows do not cost more per run — the inverse of Zapier’s per-task and Make’s per-operation meters. This is a deliberate value-metric choice: n8n optimizes for a unit the buyer can predict rather than one that maximizes per-action revenue.

2. Unlimited users, workflows, and steps on every paid plan. Since the August 2025 repricing, headcount and workflow count never move the bill; only the monthly execution quota and governance features change across tiers. n8n explicitly deleted the active-workflow caps it had carried since 2023 because they discouraged teams from building.

3. Free, open self-hosted Community Edition on the same codebase. n8n competes against its own free tier, monetizing managed hosting, governance, and support rather than core capability — a fair-code variant of the open-core model that shifts the pricing job from capability to operational convenience.

4. Two AI meters, and only one of them can ever bill you. Because one workflow run is billed once no matter how many model calls, tool invocations, or branches it spawns, an AI agent that loops through dozens of LLM calls inside a single workflow still costs one execution — a structurally different bet from the per-call token meters that make AI-agent workflows a cost monster on usage-priced platforms. The second meter, added on 2026-07-22, is stranger: an AI Assistant credit allowance of 2,300/month on Starter and up to 13,700 on Pro that refreshes monthly, never rolls over, and — per n8n’s own FAQ — cannot be topped up at any price. Most AI-era SaaS turns a credit pool into a revenue line; n8n built one it refuses to sell, so running low routes the buyer to an upgrade rather than to a bill.

5. Self-host removes the meter entirely. Run the Community Edition and you pay nothing to n8n per execution — the inverse of the cloud model. That makes n8n one of the few automation platforms where the cheapest-at-scale option is to opt out of the vendor’s billing meter altogether and absorb the infrastructure cost yourself.

Strengths & weaknesses

Billing UX : Named controls on the n8n pricing and cloud surfaces

• Monthly / Annually toggle — a switch on the pricing page swaps the displayed per-month price between annual (labelled “Annually (Save 17%)”) and monthly cadence (Starter $20↔$24, Pro $50↔$60, Business $800↔$960). The same toggle appears on the Start-up plan page — where the advertised $333 sits on the annual side of the switch, not the monthly one, so it is not directly comparable to the $24 / $60 / $960 headline figures on the cloud cards.
• Execution-volume selector on the plan card — the Pro and Business cards render their included-executions box as a dropdown, so a buyer steps the quota up from the 10k / 40k headline bucket without leaving the pricing page, and the card price re-renders with it (Pro’s 50,000-execution step is $120/mo annual · $144/mo monthly). Starter’s 2.5k box is fixed (no dropdown), and Enterprise shows a “Custom number of workflow executions” control instead.
• Cloud admin dashboard — the n8n Cloud admin surface manages instance version, timezone, concurrency, ownership/username, and workflow downloads (per the cloud docs).
• Insights dashboard — per-tier execution, failure-rate, and time-saved metrics (7 days on Pro, 30 on Business, 365 on Enterprise) for tracking usage and demonstrating ROI. Annual Business subscribers also get a proactive outreach: “the n8n team will reach out to you when you are close to reaching 80% of your annual quota,” plus a weekly execution-count email across all instances carrying the license key.
• AI Assistant credit balance — a monthly, non-purchasable allowance (2,300 Starter / up to 13,700 Pro) that “refreshes every month” with no rollover; there is no top-up control, so the only lever is a plan upgrade.
• Free-trial entitlement caveats — Starter and Pro trials need no credit card and grant “access to all Pro features,” but run under Starter plan limits (5 concurrent executions, 320MiB RAM, 10 millicore burstable CPU, 2,500 executions) and only 800 AI Assistant credits, “which expire when the trial ends.” The Business free trial is 14 days, full-feature, and does require a credit card.
• Self-service DocuSign contract for invoice / wire billing — Business (annual) and Enterprise can pay by invoice or wire; n8n publishes separate self-service DocuSign web forms for € payment and for $ payment, and the license key issues once the contract is signed. Lower tiers run self-serve via Paddle checkout.
• Cancellation controls — monthly commitments cancel any time and end at the close of the current billing cycle; annual commitments end at the close of the annual term, and “n8n does not offer pro-rata refunds.”

Strategic wins : Why the execution metric and free self-host worked

1. Choosing executions as the single value metric

By billing on completed workflow runs rather than steps or seats, n8n made its pricing legible and complexity-neutral — and the right value metric is the one buyers can predict, not the one that maximizes per-action revenue. The August 2025 repricing doubled down by deleting the workflow and seat meters entirely, leaving one dimension. That contrast against per-resolution and per-seat peers is worth comparing across the pricing blueprint.

2. Monetizing hosting and governance, not the core engine

A free open Community Edition de-risks adoption while paid cloud and the self-hosted Business license capture teams that need governance or would rather not run infrastructure. The paid story is operational convenience, SSO/Git/environments, and support — not raw capability — which is the open-core variant of usage-based packaging that lets a free tier feed the funnel instead of cannibalizing it.

3. Bundling unlimited seats to drive whole-org adoption

Because users and workflows are unlimited on every paid plan, automation can spread across a company without per-seat procurement friction, and n8n recovers the value on the execution dimension as usage grows. Removing seat-counting is a deliberate bet that whole-org adoption plus execution growth beats per-seat revenue — the same land-and-expand logic explored for usage-based SaaS.

4. Riding the AI-agent wave to reprice the category

n8n repositioned from “Zapier alternative” to AI-agent orchestration just as agents went mainstream, and its execution meter happens to absorb agent fan-out cleanly: a loop of dozens of model calls inside one workflow is still one execution. That story — plus 10× revenue growth — underwrote a $180M Series C at a $2.5B valuation in October 2025, validating execution-metered automation as an AI-era pricing model.

5. Scaling the AI entitlement instead of selling it

On 2026-07-22 n8n raised the bundled AI allowance ~46× on Starter and ~91× on Pro and simultaneously declined to sell top-ups — the FAQ says plainly that “upgrading to a higher plan is the way to get a larger allowance.” That converts AI consumption into a tier-differentiator instead of a variable charge, which protects the thing n8n actually sells: a bill that moves on one predictable axis. The cost is real (heavy Assistant users subsidise light ones, and the company forgoes the credit revenue most peers now book), but it keeps the AI feature from re-introducing the bill unpredictability the execution meter was designed to avoid, and it turns a soft feature limit into an upgrade prompt rather than an invoice surprise — the same entitlement-as-packaging logic that keeps credit-based billing legible.

Areas to improve : Surfacing the hidden Pro rung, pricing the Business ladder, and credit opacity

1. Surface the mid tier n8n already sells

The most-cited complaint after the 2025 repricing is the 16× step from Pro ($50, 10k executions) to Business ($800, 40k) — and the 2026-07-22 capture shows the mid tier already exists. Pro’s on-card selector steps to 50,000 executions at $120/mo billed annually: 5× the quota for 2.4× the price, and more executions than the Business bucket for a seventh of the money. The defect is discovery, not packaging. A buyer comparing four headline cards sees $50 → $800, the comparison table lists default buckets only, and the $120 rung appears solely to someone who opens a dropdown they have no reason to suspect. Rendering the 50k rung as a visible price on the card — or as a second column in the comparison table — would defuse the loudest objection to n8n’s pricing without changing a single number, and pairing it with thresholding and alerting as a team nears 10k would put the cheap step in front of them at the moment it matters.

2. Publish the Business ladder, not just its entry rung

Pro’s selector at least re-prices itself when a buyer operates it. Business’s does not resolve to published figures: the card runs a 40k→500k ladder whose higher rungs are quoted per-currency at checkout and never appear as static numbers, and Enterprise remains fully “contact sales.” A team that outgrows even Pro-50k therefore hits an unpriced wall exactly where the deal gets large enough to need a budget approved — and, because Business is self-host-only, that wall arrives with an infrastructure decision attached. Indicative volume bands for the Business rungs would let buyers size a year ahead instead of opening a sales cycle to learn a list price, the way the better-documented usage-invoicing playbooks recommend.

3. Finish the at-quota story below Business

n8n answered half of this on 2026-07-22: the FAQ now states that workflows keep running past quota, that sales reaches out first, that unaddressed overages are auto-invoiced 45 days later, and that Business overage costs 4,000 EUR per extra 300,000-execution bucket. Two gaps remain. That bucket is 30× an entire Pro quota, so the smallest possible overage charge is far larger than the tier most teams overshoot from — it smooths the top of the range while leaving the rungs where real teams actually overshoot — a Starter at 2,500 or a Pro at 10,000 — with no priced remedy at all. And the policy is scoped to Business and Enterprise only, so exactly those Starter and Pro customers still cannot read what happens when their bucket empties, which is where bill-shock anxiety actually lives. Publishing a Pro-scale overage rate — and pairing it with threshold alerting in the insights dashboard before the limit hits — would close the question for the tiers that ask it most.

4. Give the AI credit allowance a meter buyers can read

The new AI Assistant allowance is generous but unreadable: n8n publishes the balance (2,300 on Starter, 5,700 or 13,700 on Pro) yet says consumption “depends on the size and complexity of what you’re building,” so a buyer cannot translate credits into work done before subscribing. Because credits cannot be bought and do not roll over, an exhausted balance has exactly one remedy — a plan upgrade — which makes the burn rate a purchase-relevant number, not a detail. A published per-action credit cost and an in-product balance indicator with usage-grant tracking would let teams size the allowance up front instead of discovering it mid-month. The “up to 13,700” phrasing on the Pro card compounds the problem: the true figure is 5,700 or 13,700 depending on plan size, and only the FAQ says so.

Key takeaways

1. Pick one value metric and commit. n8n puts its entire price gradient on monthly executions, keeping the model legible where per-step and per-task rivals feel arbitrary. The 2025 repricing went further and deleted its other meters (workflows, seats) rather than stacking them.
2. Removing a meter can be a bigger move than adding one. n8n’s most consequential pricing change was subtractive — killing active-workflow caps — because the caps were discouraging the exact building behavior the product depends on. Constraints that suppress adoption cost more than they collect.
3. Unlimited seats can be a feature, not a giveaway — and so can an un-sellable credit pool. Removing seat counting lets automation spread across an org without procurement friction, with cost recovered on the usage dimension as workloads grow. n8n ran the same play on AI in July 2026: it scaled the bundled Assistant allowance ~46×/91× and explicitly refused to sell top-ups, so the entitlement recruits upgrades instead of generating invoices.
4. A free self-hosted tier can be a moat, not a leak. Open-core (here, fair-code) lowers adoption risk and feeds the funnel into paid managed hosting and governance, as long as the paid value story is operational rather than capability-based.
5. A cliff can be an artifact of how you display tiers, not how you price them. n8n’s most-cited pricing complaint is a 16× Pro→Business jump — yet a $120 rung at 50,000 executions sits inside the Pro card, cheaper than Business and larger than its bucket. The ladder was fine; the four headline cards hid the middle of it. Buyers price you off what renders by default, so a rung behind a dropdown is, commercially, a rung that does not exist.

UBP implications

1. Complexity-neutral metering reduces bill anxiety. Counting a whole workflow as one unit — regardless of internal fan-out — shows that the “right” usage unit is the one buyers can predict, not the one that maximizes per-action revenue.
2. Where you put the overage decides who it protects. n8n does publish an overage — but only on Business, and only in 300,000-execution buckets at 4,000 EUR, roughly 30× a whole Pro quota. Granularity, not existence, is what determines whether an overage smooths a cliff or merely bills the customers already past it.
3. Not every meter has to bill. When the engine is free to self-host, pricing must monetize hosting, governance, and support rather than capability — and n8n extends that logic to AI by shipping a credit allowance it deliberately refuses to sell. A metered entitlement can do packaging work (tier differentiation, upgrade pressure) without ever becoming a revenue line, which is the road not taken by most AI-era credit pools.

Sources

• n8n pricing page (accessed 2026-07-22)
• n8n Startup plan (accessed 2026-07-22)
• n8n for Enterprise (accessed 2026-07-22)
• n8n Cloud docs (overview) (accessed 2026-07-22)
• n8n docs — AI Assistant (preview) and credits (accessed 2026-07-22)
• n8n blog — new pricing explained (2025-08-07) (accessed 2026-06-02)
• n8n blog — Sustainable Use License announcement (accessed 2026-06-02)
• n8n blog — Series C announcement (accessed 2026-06-02)
• n8n Sustainable Use License (docs) (accessed 2026-06-02)
• n8n pricing page — JS-rendered USD read (independent second-source price validation) (accessed 2026-06-02)
• n8n pricing & startup-plan pages — live cadence re-validation (EUR geo-render confirms per-currency price points; $333 Startup verified as “/mo, billed annually”) (accessed 2026-06-05)
• n8n pricing page — JS-rendered USD read + expanded FAQ (AI Assistant credit allowances, Business overage rate, free-trial entitlements) (accessed 2026-07-22)
• n8n pricing page — plan catalog behind the on-card execution-volume selector (Pro 10k and Pro 50k options, per-plan concurrency and AI-credit tooltips) (accessed 2026-07-22)
• n8n Cloud product catalog at Paddle, n8n’s merchant of record — USD price and billing interval per plan SKU (n8n Cloud Starter / Pro-1 / Pro-2 and their yearly SKUs) (accessed 2026-07-22)
• n8n Start-up plan page — independent US-geo re-render confirming “$333/mo, billed annually” as an annual-cadence price alongside the page’s own Monthly/Annually toggle (accessed 2026-07-22)
• n8n docs — AI Assistant credits during preview (“To get more credits during Preview, upgrade your plan”) (accessed 2026-07-22)

Bottom line

n8n prices workflow automation on a single, complexity-neutral metric — monthly workflow executions — with unlimited users and workflows on every paid plan, a free open self-hosted edition, and sales-led Enterprise on top. The July 2026 packaging change tested that discipline and confirmed it: n8n added a second per-plan meter for AI Assistant credits, scaled it ~46×/91×, and then refused to sell it, so the allowance sorts buyers into tiers without ever moving the bill. It remains one of the cleanest examples of putting an entire price gradient on one usage dimension.

Want to compare n8n against other workflow-automation pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

AI Assistant credit allowances replace AI Workflow Builder credits

Jul 2026

The bundled AI entitlement on every cloud plan card was renamed from 'AI Workflow Builder credits' to 'AI Assistant credits (preview)' and the monthly allowance jumped from 50 to 2,300 on Starter and from 150 to 'up to 13,700' on Pro (5,700 or 13,700 depending on plan size). Enterprise's 1,000 cloud-only credit line disappeared — Business and Enterprise now read 'AI Assistant coming soon' and the comparison table shows N/A. Credits reset monthly, do not roll over, and cannot be bought separately, so a larger allowance means an upgrade. The comparison table also gained a saved-execution storage row (2.5 GB / 25 GB / 50 GB). Plan prices were unchanged.

Execution-tiered cloud pricing

Jun 2026

Cloud plans Starter $20, Pro $50, Business $800 (monthly equivalents when billed annually), each defined by an included monthly workflow-execution quota of 2.5k / 10k / 40k. Annual billing saves 17%. Enterprise is custom-quoted; Startup plan is $333 (50% off Business).

$180M Series C at $2.5B valuation — AI-orchestration repositioning

Oct 2025

Accel-led $180M Series C (with NVIDIA's NVentures, Meritech, Redpoint) valued n8n at $2.5B, total funding $240M, on 6× user and 10× revenue growth. n8n repositioned from 'Zapier alternative' to AI agent orchestration. 235-point HN thread on 2025-10-09. (n8n blog /series-c.)

Repricing: active-workflow limits removed; Business plan added; execution-only metric

Aug 2025

n8n removed active-workflow limits on every plan, making users, workflows, and steps unlimited, and put the entire price gradient on monthly executions. A new Business plan (esp. for self-hosted teams) slotted between Pro and Enterprise. n8n framed the old workflow-based caps as discouraging enterprise builders from experimenting. (n8n blog, 2025-08-07.)

Starter execution quota lowered to 2.5k; annual toggle at 20% off

Jan 2024

Wayback snapshots (2024-01, 2024-10) show Starter at 2,500 executions (down from 5k), Pro at 10k, and a Monthly/Annual toggle advertising 20% annual savings. Still a 3-tier Starter/Pro/Enterprise layout with active-workflow caps; no Business tier yet. Dollar figures did not render in the archive (unknown).

Cloud paid tiers: Starter / Pro / Enterprise with active-workflow caps

Jul 2023

Wayback snapshot (2023-07) shows three cloud tiers — Starter (5k executions, 5 active workflows, single user), Pro (10k executions, 10 active workflows, unlimited users), Enterprise (custom). Pricing combined an execution quota with an active-workflow cap and, on Starter, a single-user limit. Dollar figures did not render in the archive (marked unknown).

Pricing page is self-host first; Cloud in early access

Jun 2022

Wayback snapshot (2022-06) shows the pricing page as a Desktop & Self-hosted / Cloud toggle: Desktop App (Free) and Self-Hosted (Free) are the headline offers; managed Cloud sits behind a separate tab in early access. No paid cloud tiers are surfaced as the primary view.

Apache 2.0 → Sustainable Use License (fair-code)

Mar 2022

n8n replaced its Apache 2.0 + Commons Clause license with the new Sustainable Use License — source-available, based on Elastic License 2.0 — and coined the term 'fair-code'. The self-host engine stays free; commercial hosting-as-a-service is restricted. (n8n blog, 2022-03-17.)