Relevance AI Pricing in 2026: The Public Numbers Are Gone

What does Relevance AI cost in 2026?

Relevance AI pricing is not public. I loaded relevanceai.com/pricing on 17 August 2026 and the page contains one tier, Enterprise, and zero dollar amounts. The only call to action is "Talk to sales".

That matters more than it sounds, because most of the pricing articles ranking for this question quote confident numbers. Those numbers came from an earlier version of the page. They are not wrong so much as expired, and they no longer agree with each other.

Only the last row is currently verifiable. Treat the rest as archaeology.

What the Enterprise tier actually includes

The live page does publish the feature list, even without prices. Enterprise covers custom Actions and Vendor Credits, unlimited agents and tools, unlimited users and projects, unlimited workforces, 2,000+ integrations, calling and meeting agents, enterprise triggers, agent evaluations, A/B testing and analytics, SSO with RBAC and audit logs, and a dedicated account manager.

That is a complete enterprise feature set. The gate is commercial, not technical.

How Relevance AI pricing works: Actions and Vendor Credits

Even without prices, the billing model is documented and worth understanding, because it is a dual meter rather than a single number.

• Actions. One Action is one tool run. This is the platform fee, and it resets each billing cycle.
• Vendor Credits. These are pass-through model costs, billed at cost with no markup. They roll over indefinitely while you stay subscribed.

The dual meter is genuinely more transparent than a single blended credit, because you can see what you are paying the platform versus what you are paying OpenAI or Anthropic. Most competing platforms blend both into one opaque credit, so you cannot tell whether a price rise came from the vendor's margin or from the underlying model. Relevance AI also supports bringing your own API keys, which lets you bypass Vendor Credits entirely and pay your model provider directly, and that is the single biggest lever on total cost if you already hold provider pricing.

Reported top-up rates, from secondary sources rather than the vendor, sit around $80 per 1,000 Actions and $20 per 10,000 Vendor Credits (Automation Atlas). I could not confirm these on any Relevance AI page, so treat them as indicative.

The cost trap nobody advertises

One characteristic of Action-based billing deserves a flag: failed and test runs consume Actions. Debugging a misconfigured agent draws from the same meter as production traffic, so the build phase costs real money and the learning curve is billable (11x).

That source is a competitor, so weigh it accordingly, but the mechanism follows directly from the documented model. If an Action is a tool run, a failed tool run is still a tool run.

Why vendors hide pricing

Removing public pricing is a deliberate move and usually signals one of three things: a shift upmarket where deals are negotiated rather than self-served, pricing volatile enough that publishing it creates churn, or usage patterns varied enough that one number would mislead.

None of those are scandalous. But there is a practical cost to buyers. You cannot budget, you cannot compare, and you cannot get a number without entering a sales cycle. For a team evaluating four platforms in a fortnight, that alone can knock a vendor off the list.

The contrast is visible across the category. Gumloop pricing publishes a $37 entry point and its orchestration fee. Lindy AI pricing publishes tiers. n8n pricing publishes tiers. Relevance AI, Glean and Moveworks all require a conversation.

What to ask if you do talk to sales

Because there is no list price to anchor against, the questions matter more than usual:

1. What is the annual floor, and is there a minimum commitment?
2. How many Actions does a realistic month of my workload consume? Ask them to model it, not to quote a tier.
3. Do failed and test runs count against Actions?
4. What happens when I exhaust Actions mid-cycle? Hard stop, or overage billing?
5. Can I bring my own model keys from day one, or is that gated to a higher tier?
6. Do unused Actions roll over, or only Vendor Credits?

Question 2 is the one that decides the bill. Everything else is terms.

How this compares to per-seat pricing

Coworker AI publishes its pricing: Pro at $29.99 per user per month, Max at $149.99, and Enterprise custom. You can budget it before you talk to anyone.

Per-seat pricing has a real weakness worth stating plainly. If you have many light users, you pay for capacity you do not use, and a consumption model would be cheaper. What per-seat buys you is a number you can forecast, and an agent that runs a thousand times costs the same as one that runs ten times.

Coworker connects to 50+ enterprise tools and includes an agent builder, so the capability comparison is fair. The pricing philosophies are simply opposite. For the wider view, see our enterprise AI pricing comparison and how much enterprise AI should cost per user.

Related depth: AI agent orchestration platforms, best AI workflow automation platforms, and best enterprise AI workflow automation.

Book a demo if you want a number without a discovery call first.

Frequently asked questions

How much does Relevance AI cost?

There is no published price. As of August 2026 the Relevance AI pricing page lists only an Enterprise tier with a "Talk to sales" button and no dollar figures. Any specific monthly price you find in a third-party article reflects an earlier version of the page.

Does Relevance AI still have a free tier?

The live pricing page does not show one. Documentation and secondary sources describe a free allocation of 200 Actions per month plus a one-time grant of Vendor Credits, but this is not advertised on the pricing page, so confirm it directly before planning around it.

What is the difference between Actions and Vendor Credits?

An Action is one tool run and represents the platform fee. A Vendor Credit is the underlying AI model cost, passed through without markup. Actions reset each cycle; Vendor Credits roll over while you remain subscribed.

Do failed runs cost money in Relevance AI?

Reportedly yes. Because billing counts tool runs rather than successful outcomes, testing and debugging draw from the same Action pool as production. Confirm the specifics with their team, as this comes from third-party reporting rather than published documentation.

Can I use my own OpenAI or Anthropic keys?

Yes. Relevance AI supports bringing your own API keys, which bypasses Vendor Credits so you pay your model provider directly. This is the main lever for controlling model spend if you already have provider pricing.

Why did Relevance AI remove its public pricing?

The company has not said. Removing published pricing typically accompanies a move upmarket toward negotiated enterprise contracts. The practical effect for buyers is that comparison and budgeting now require a sales conversation.

Related reading

• Gumloop pricing
• Lindy AI pricing
• Enterprise AI pricing comparison
• Glean pricing
• Best AI workflow automation platforms