The short answer: most AI agents cost somewhere between free and about $200 a month for an individual, and between roughly $1 and $2 per resolved conversation for support agents. The sticker price matters less than the meter. Almost every agent bills on one of four meters (seats, outcomes, minutes or credits), and the meter decides whether your bill stays flat or grows with usage. This guide is built from the pricing data on our agent profiles, which we re-verified in October 2026.

Quick picks

• Cheapest serious start for a team: n8n, free self-hosted, cloud from $24/month.
• Cheapest AI coding assistant: GitHub Copilot, free plan, Pro $10/month.
• Clearest outcome pricing for support: Intercom Fin, $0.99 per outcome plus seats, or $49.50/month minimum standalone.
• Lowest voice platform fee: Vapi, $0.05/min platform fee plus model and telephony costs.
• Most predictable no-code agent builder: Lindy, from $29.99/month, but credits don't roll over.

The four pricing models

1. Per seat

You pay per human user per month. This is how coding tools and helpdesks bill. Cursor runs Hobby (free), Pro $20, Pro+ $60, Ultra $200 and Teams $40 per user. GitHub Copilot runs Free, Pro $10, Pro+ $39, Max $100, Business $19 per user and Enterprise $39 per user. Zendesk Suite starts at $55 per agent per month and an AI Copilot add-on is another $50 per agent.

The catch: seat pricing doesn't fall when the AI does more work, and heavy users on "unlimited" tiers often hit usage caps or move to a pricier plan. Check the limits, not just the headline.

2. Per outcome or resolution

You pay only when the agent finishes the job. Intercom Fin charges $0.99 per outcome on top of Intercom seats ($29, $85 or $132 per seat per month by tier). Fini is $0.89 per resolved ticket on Growth, falling to $0.69 on Scale and $0.49 on Enterprise. Zendesk quotes roughly $1 to $2 per resolution, Decagon is custom (we estimate about $1.50), and Ada is $1 to $3.50 per resolution, shifting toward conversation pricing, with annual contracts from about $30K a year.

The catch: "resolution" is defined by the vendor. Ask what counts, whether a customer who abandons counts, and whether there is a monthly minimum. Salesforce's Agentforce shows how much the unit matters: $2 per conversation, or $500 per 100,000 Flex Credits. For a deeper look at this category, see our best AI customer support agents ranking.

3. Per minute

Voice agents bill by call time. Bland AI is $0.14/min on Start, or $299/month plus $0.12/min on Build. Retell AI has no platform fee and starts at about $0.07/min, plus $8 per concurrent call per month. Vapi charges a $0.05/min platform fee and gives $5 in free credits.

The catch: a platform fee is rarely the whole bill. Speech, language-model and phone costs stack on top, so a quoted "$0.05 a minute" is a floor, not a forecast. Test with a realistic call script before you commit. We compare these in best AI voice agents.

4. Credits and usage

You buy a pool of credits and each action burns some. Lindy starts at $29.99/month (Plus), $99.99 (Pro) and $199.99 (Max), and calls cost extra. Relevance AI has a free plan, Pro at $19/month billed annually, and Team at $234/month billed annually. Artisan offers a free tier with 300 credits a month, then $250/month (Intern) or $600/month (Employee) on annual plans. n8n bills by workflow execution: free self-hosted, Starter $24, Pro $60 and Business $800 per month on cloud.

The catch: credits expire, burn faster on AI-heavy steps, and are hard to forecast. See best AI workflow automation agents for how this plays out.

Where headline prices mislead

• Contract-only tools. 11x lists about $3,750/month on an annual contract, rising to $6,500 to $8,500 with voice. That is closer to a salary line than a software subscription, and our profile scores it 4.8, partly on value.
• Enterprise "custom" pricing. Decagon, Sierra, Glean and Harvey don't publish rates. Budget for a sales cycle, not a credit card.
• Model costs outside the plan. API prices can change under you. In August 2026 DeepSeek moved to peak and off-peak rates after years of flat pricing, a reminder that any agent built on pay-per-token models inherits that risk.
• Minimums and add-ons. Fin's standalone minimum, Zendesk's Copilot add-on, and Retell's concurrency fee are all cheap individually and easy to miss.

Which pricing model should you choose?

• If your volume is low or unpredictable, prefer outcome or pay-as-you-go pricing and avoid annual contracts.
• If your usage is steady and heavy, seats or flat plans win, as long as the usage caps fit.
• If you're a developer on a budget, start with GitHub Copilot Pro at $10 or Cursor Pro at $20, and see best AI coding agents.
• If you can self-host, n8n removes the meter entirely; you pay for hosting and your own time.
• If finance needs a hard monthly cap, choose flat tiers and confirm what happens at the limit.

Whatever you choose, pilot one narrow workflow for a month, record the real cost per finished task, and compare that, not the sticker price. If you're not sure an agent is the right tool, read choosing the right AI agent.

Next step

Open the Intercom Fin profile if you run support, the Vapi profile if you're building voice, or the n8n profile for the cheapest automation start. Browse every plan side by side in the customer support, voice and automation categories.

Frequently asked questions

How much do AI agents cost in 2026?

Individual tools run from free to about $200 a month (Cursor Ultra and Devin Max are both $200). Support agents typically bill around $1 to $2 per resolved conversation, and enterprise sales agents like 11x cost thousands per month.

What is the cheapest AI agent?

Several are free: n8n self-hosted, Aider, Jan and Ollama for local models. Among paid tools, GitHub Copilot Pro at $10/month is one of the cheapest coding assistants.

What is outcome-based pricing?

You pay per completed result, such as a resolved support ticket, instead of per seat. Fin charges $0.99 per outcome and Fini $0.89 per resolved ticket. Ask how each vendor defines a "resolution".

Why is AI agent pricing so hard to compare?

Vendors use different units (seats, resolutions, minutes, credits, executions), and some also add model, phone or add-on fees. Convert each quote into cost per finished task before comparing.

Are AI agent prices stable?

No. Vendors regularly change plans, credits and limits, so check the vendor's current page before buying. We re-verify the figures on each profile and show the review date there.

Methodology note: prices come from the pricing sections of our agent profiles, last reviewed October 2026. Enterprise estimates (such as Decagon's) are approximations, not published rates.