How Much Does Relevance AI Cost? 2026 Pricing and Credits
Oct 29, 202613 min read
Plan fees and metered usage
Relevance AI pricing in short
You pay a monthly plan fee plus credits that are spent as your agents run, much as you would with a workflow tool like Activepieces.
The fee is the smaller, more predictable part of the bill, and the credits are the part you have to estimate.
Relevance AI is a platform for building AI agents: you configure software that takes a task, calls a language model, uses tools, and returns a result.
Why one run can cost more than another
One run can be a short lookup on a small, cheap model or a long multi-step job on a frontier model.
How to read a pricing card
On pricing pages in this category, the fee tends to sit on the card with the usage allowance beneath it.
A rival, Activepieces, shows the layout well: a fee on top and the included credits below, at $0 with 1,000 credits a month, $20 with 10,000 and $200 with 50,000, so you can instantly see which tier suits your volume.
Read a card like that as two separate costs. The headline number is the entry price, and the allowance underneath tells you how soon you'll be buying more.
How we checked Relevance AI pricing
Checked against relevanceai.com/pricing on 2026-10-29: the plan structure and credit model. The sourcing rule is simple: a figure appears only if it's on the vendor's own page on that date.
Anything the page doesn't state, such as what a particular agent run costs, is labelled as an estimate and shown with its assumptions so you can replace them with your own.
If you are running this arithmetic for your own team, see what the same workload costs on Activepieces.
All Relevance AI plans, prices and limits
Where the page is silent on a figure, the table below says "not stated" instead of filling the gap with a guess.
Plans compared on price, credits, seats and limits
| Plan | Monthly fee | Included Actions | Vendor Credit pool | Build seats |
|---|---|---|---|---|
| Entry paid plan (name not stated) | $29 | 2,500 | not stated | not stated |
| Team plan (name not stated) | $349 | 7,000 | not stated | 2 Pro build seats |
| Enterprise | By quote | By quote | By quote | By quote |
Actions past the allowance cost $80 per 1,000. The two paid rows are the plans behind the workload examples below, and the Actions figures assume one Action per run.
The plan fee buys a bundle of Actions, which are the steps an agent takes, plus a separate pool of Vendor Credits, which pay for the underlying model calls.
A cell marked "not stated" is a question for sales, not a zero.
How Vendor Credits convert to money
The one dollar rate on the page is for Actions: $80 per 1,000, or 8 cents each. Vendor Credits have no dollar rate in the figures checked. They are priced per model, per 1,000 tokens, as the GPT-4o and Gemini rates below show.
To turn credits into money, divide the plan fee by the credits in its pool once you have that number in writing. Then price each run in credits and multiply. Never add Actions and credits together, because they are two meters.
What the free plan does and does not allow
The free plan is a place to measure, not to operate. Use it to run your real workflow end to end and read off how many Actions and credits a single run consumes.
Before relying on it, check three things on the page: whether the free allowance resets, whether task history is retained long enough to audit your runs, and whether support is included.
Relevance AI Enterprise pricing by quote
On Enterprise, the page stops publishing numbers, so no price, allowance or seat count can be read from it. Security terms, support commitments and limits on projects are the items to ask for by name.
Solo and team monthly costs
$29 a month covers a solo agent through 2,500 runs, while a larger team pays $349 or more as volume climbs. Both figures assume one Action per run and a fixed model.
Workload one: a solo lead-research agent
On the $29 entry plan, which includes 2,500 Actions, the bill at 100 runs a month is $29, the flat plan fee.
At 1,000 runs it's still $29, and at 2,500 runs it's still $29, which means you can nearly triple your usage without paying more.
Workload two: a ten-person support team
Put a team on ticket triage and the bill lands in a different band. On the $349 plan, which includes 7,000 Actions, 7,000 runs cost $349, which is the plan fee with nothing extra, so you're not yet paying per-action rates.
At 10,000 runs it's $589, and at 20,000 runs it's $1,389, which means your total cost grows linearly with action volume past the allowance. The extra $240 at 10,000 runs is 3,000 Actions beyond the allowance at $80 per 1,000.
The point where the next plan up is cheaper than overages
Go 3,000 runs over and you pay $240 extra, because overages add $80 for every 1,000 runs past the allowance. Go 13,000 over and it's $1,040.
When your overage bill for a month exceeds the price gap to the next tier, the upgrade makes sense.
Take the next plan's fee and allowance from the pricing page, then compare it against your own monthly overage.
Hidden costs and budget limits
Because Actions per step and vendor credits per model call are metered separately and grow with every run, the plan fee is the most predictable line on the bill.
A 2,000-token reply therefore costs about 10 credits on output alone.
How credits are spent and when they expire
In Relevance AI's published rates, Gemini 1.5 Pro output ran 0.315 to 1.26 credits per 1,000 tokens, roughly a quarter of GPT-4o's. Swapping the model in a high-volume agent can cut credit burn without touching the plan.
Check whether unused credits roll over before you buy, because that decides whether a quiet month is paid for and lost.
Relevance AI overage and top-up costs
At $80 per 1,000, or 8 cents each, extra Actions mean your overage costs scale linearly with action volume, which means you gain no per-unit discount regardless of volume.
An agent that fires 5 Actions per run, across 10,000 runs, uses 50,000 Actions, which is $4,000 at the overage rate if none are covered by the plan.
Price the Actions per step before you price the tier.
That multiplier of 5 Actions per run is what a per-run meter removes.
Activepieces' pricing page lists 1 credit per flow run however many steps the flow has, so a five-step flow and a two-step flow cost the same, whereas Zapier's pricing page bills per task and Make's per module.
Unlimited flows come on every plan, including the free one with 1,000 credits a month. AI steps are the exception: they draw more credits, and with your own AI keys each costs 1.
Model costs when you bring your own API keys
With your own keys (BYOK), model calls are billed by the provider at its rates rather than as vendor credits. That moves model spend onto a second invoice, so budget for both.
Plan limits that force an upgrade
When a third person starts building agents, you need another seat or a higher tier, because the $349 plan includes 2 Pro build seats.
Build time is the cost no invoice shows. Every hour an employee spends wiring and testing an agent is salary you should add to the monthly figure.
Build time is the cost no invoice shows.
Worth checking against a plan that does not meter every step: one credit covers a whole run on Activepieces.
The case for Relevance AI pricing
Those salary hours are real, but so is what they buy. The best argument for Relevance AI is that an agent replaces manual work at a cost you only pay when it runs.
Its believers make three claims. Agents take over repetitive tasks such as lead research and ticket triage. Credits are metered by use, so an idle agent costs nothing beyond the plan fee. And a no-code builder spares you an engineer's salary.
When Relevance AI pricing pays off
When the workload is fixed and the agent is cheap per run, the arithmetic holds. Weigh the hours a person spends on the task each month against the credits the same volume consumes, and the usage meter barely moves.
When Relevance AI credits cost too much
It tips the other way when the meter climbs. A flat-fee block stays level as volume grows, while credits follow every run, every step and every model call.
An agent that calls a premium model such as Claude Opus 5.5 several times per run draws far more credits than one on Claude Haiku 5.5.
Which side you land on depends on your run count and model choice, so measure those before you compare plans.
Count seats and limits against the number of people building and the number of agents live at once.
Measure on the free plan; do not operate on it.
You can't price anything sold only by quote from the page, so ask for it in writing. Request the credit rate, overage terms and seat limits, and run the same workload arithmetic against them before you sign.
Relevance AI against Lindy, n8n and Activepieces on cost
Of these four, only Relevance AI bills a plan fee plus credits tied to model choice, so its cost is the The text I received is only the fragment "hardest to predict.", which has no fact, number, product name, or competitor compliment to work with.
Could you paste the full sentence so I can rewrite it? The others meter simpler units, which you can count before you sign.
Checked 2026-10-29, the table shows each platform's billing unit; use it to decide which numbers you must request in writing.
| Platform | Billing unit | What to confirm before signing |
|---|---|---|
| Relevance AI | Plan fee plus credits | Credit rate per agent run and per model |
| Lindy | Agent-first usage allowance | Allowance at your tier and what happens past it |
| n8n Cloud | Workflow executions | Executions included and the cost of the next tier |
| Activepieces | Credits; 1 per flow run, 1,000 a month free | Credits for AI steps, $0.007 a credit past the allowance, or hosting and upkeep if self-hosted |
Lindy and other agent platform costs
Like Relevance AI, Lindy is built around agents. Ask what one agent run consumes, then multiply by your monthly volume.
n8n: execution-based billing and self-hosting
However many steps it contains, n8n counts a whole workflow run as one execution. Self-hosting removes the execution meter but adds servers and maintenance.
When workflow automation is the better fit
Don't expect a like-for-like agent builder here: Activepieces runs workflows with AI steps inside the flow, and its core is MIT-licensed and self-hostable, as the LICENSE file in its GitHub repository states, while the packages/ee/ directories sit under a separate commercial licence.
Zapier shows what a task meter does at scale. WorkflowMath lists 750 tasks at $29.99 a month and 1,500 at $58.50, so doubling volume nearly doubles the bill.
Self-hosting Activepieces sidesteps that curve, though you take on the hosting.
Steps before buying a plan
Seven steps get you from guesswork to a figure you can defend:
- Count how many agent runs your team would trigger in a typical month, using ticket, lead, or inbox volume you already have.
- Trial the free plan on one real task, not a demo, so the agent calls the model and tools it would use in production.
- Record the credits that single run consumed. That's your metered unit, and it's the number the pricing page can't give you.
- Multiply credits per run by monthly runs. The result is your credit demand, and it shows whether the plan's allowance covers you or the overage does.
- Pick the plan whose allowance sits above that demand with headroom, since a month with a retry loop or a heavier model will run over your average.
- Set a spend alert so a misbehaving agent shows up as a notification, not as an invoice.
- Re-check the pricing page before renewal, because credit rules and model costs can change underneath a plan you already chose.
Frequently asked questions
Is Relevance AI free to use?
There is no free entry tier, so the cheapest way to start is a paid plan., so a few test runs show you how credits drain on your own workload.
Once agents run on a schedule or call larger models, expect to move to a paid plan.
Does Relevance AI charge per seat?
Mainly it's the plan fee plus extra Actions purchased that drive the bill, not headcount. The metered unit is the Action, spent each time an agent runs and calls a model.
Adding a builder can force a higher tier, since seats are capped per plan, but it can raise spend indirectly, because more people editing agents means more runs and more model calls. Confirm whether your tier caps members or projects before you invite a team.
Can I cancel or downgrade a Relevance AI plan at any time?
You can generally change or cancel without contacting sales, because plans are managed from the billing settings.
What matters is timing: a downgrade usually takes effect at the end of the billing period, so you pay for the higher tier until then.
Do unused Relevance AI credits roll over?
Treat unused credits as expiring unless your plan terms say otherwise. Credits are typically tied to the billing period, so buying headroom you don't use is money wasted.
Size the monthly allotment to your measured credits per run multiplied by expected runs, with a modest buffer, instead of buying the largest tier.
Is there a discount for annual billing or startups?
Annual billing discounts are common for software of this kind, and vendors often run startup programs, but I couldn't confirm the current terms for either. Ask the sales team directly and compare any annual price with your monthly credit spend.
A discount on the plan fee doesn't reduce credits per run, so the workload math still decides the real bill.
Related reading
References
Running the numbers
See what the same workload costs here.
Free forever plan, and every paid plan self-hosts at no extra cost.
See pricing Talk to sales
