Enterprise agent platform pricing, compared
Last updated: November 9, 2026
Short answer: Enterprise agent platforms charge in very different units. Microsoft, Salesforce and IBM sell credits, conversations or tiered plans. Google and AWS meter compute and model tokens. ServiceNow, Kore.ai and OpenAI Frontier are quote-only. All prices below are US-dollar list prices from vendors' official pages. Contracts, taxes, model tokens and data costs can change what you actually pay.
Pricing at a glance
| Platform | Pricing unit | Published list prices | Free start |
|---|---|---|---|
| Microsoft Copilot Studio | Copilot Credits (tenant capacity packs or pay-as-you-go) | $200/month per 25,000 credits; up to 20% off with pre-purchase; internal agents included with Microsoft 365 Copilot at $30/user/month (paid yearly) | Free trial; $200 Azure credit for new Azure accounts |
| Salesforce Agentforce | Flex Credits, conversations or per user | $500 per 100,000 Flex Credits; $2 per conversation; $5/user/month user license (needs Flex Credits); $50/user/month Coworker; $125/user/month add-ons ($150 for Industries); Max Edition from $550/user/month | Salesforce Foundations, $0 |
| Google Gemini Enterprise Agent Platform | vCPU-hours, GiB-hours, storage, operations, tokens | Agent Runtime $0.085/vCPU-hour and $0.009/GiB-hour on demand; models priced separately (Gemini 3.8 Flash $0.75/$3.75 per 1M tokens through Dec 31, 2026, then $1.50/$7.50); Google Search grounding: 5,000 queries/month free, then $14 per 1,000 | Google Cloud free trial |
| Gemini Enterprise (employee app) | Per seat | Business edition from $21/seat/month; Standard, Plus and Frontline via sales or console | 30-day trials |
| Amazon Bedrock AgentCore | Per-second CPU and memory, per query, per API call | Web Search $7 per 1,000 queries; Web Answer $10 per 1,000 tool calls; Runtime, Browser and Code Interpreter billed on actual CPU and memory used; Instances = EC2 cost plus a management fee | Up to $200 in AWS Free Tier credits for new customers |
| ServiceNow AI Agents | Assists, under contract | Not published; workflow executions use 25, 50 or 150 assists by size | Demo |
| IBM watsonx Orchestrate | Monthly tier | Essentials from $578/month (4,000 MAU, 200,000 messages); Standard from $6,360/month (40,000 MAU, 2,000,000 messages); Premium by quote | 30-day trial |
| UiPath | Plan tier | Community free; Basic from $25/month; Standard and Enterprise by quote | Free Community; 60-day Standard trial |
| OpenAI Agents API / SDK | Model tokens + tools + containers | Charged at the selected model's API rates, e.g. gpt-6.1-sol $2 input / $10 output per 1M tokens, gpt-6-luna $0.10 / $0.50 | Pay as you go |
| OpenAI Frontier | Contract | Not published | Contact sales |
| Kore.ai | Contract | Not published | Demo |
| Oracle AI Agent Studio | Included with Fusion; per user or employee for custom agents | Custom AI Agents $50/authorized user/month (10-user minimum) or $2.50/employee/month (500 minimum, ERP/SCM/HCM); additional tokens $500 per 1B | Templates included with Fusion SaaS |
| WRITER | Per seat with credit limits | Starter per-seat plans (up to 5 users); Enterprise by quote | 14-day trial |
What drives the real cost
- Actions per task. On Agentforce, Salesforce's examples count 20 Flex Credits ($0.10) per action. A three-action case therefore costs about $0.30.
- Model choice. On Google, AWS and OpenAI, model tokens are often the largest line item. Promotional Google model prices change on stated dates, for example Gemini 4 Argon moves from $2/$10 to $4/$20 per 1M tokens on December 1, 2026.
- Idle time. AWS AgentCore's microVMs and Google's Agent Runtime don't charge CPU while an agent waits on model or tool responses, which matters for long-running agents.
- Commitments. Microsoft (pre-purchase), Salesforce (pre-purchase or pre-commit) and Google (Flexible Savings Plans) discount up-front commitments. Unused Salesforce Flex Credits do not roll over.
- Hidden extras. Watch for Salesforce Data 360 credits, AWS storage and data transfer, Google grounding queries, and premium-model token fees on Oracle.
Notes on sources
Prices were taken from each vendor's official pricing or documentation page on November 9, 2026. Google's Agent Runtime rates, ServiceNow's assist metering and Oracle's Custom AI Agents prices come from official pages we could not read directly with our tools; they were verified through a research summary of those pages. See the methodology for details.