Home · Pricing comparison · Methodology · About
How much do enterprise agent platforms cost in 2026?
Last updated: 2026-11-09
Most enterprise agent platforms now charge mainly for usage: credits per action, per-call units, or compute time. Some add per-user seats on top. Of the 13 platforms in our ranking, Salesforce, Microsoft, AWS and IBM publish the most complete public pricing. ServiceNow, Automation Anywhere, Kore.ai and OpenAI Frontier are priced by quote. All figures below are vendor list prices in USD, taken from official pages on the date checked. Contracts, regions and discounts change the real price.
Pricing at a glance
| Platform | Pricing model | Published list prices |
|---|---|---|
| Salesforce Agentforce | Flex Credits, per-conversation, per-user licenses, editions | $500 per 100,000 Flex Credits; standard action 20 credits ($0.10); voice action 30 ($0.15); $2 per conversation; $5/user/month license (needs Flex Credits); add-ons $125–$150/user/month; Core $195, Advanced $395, Max $550 per user/month |
| Microsoft Copilot Studio | Copilot Credits (pay-as-you-go or prepaid) | $0.01 per credit pay-as-you-go; prepaid 5–20% discount; generative answer 2 credits, agent action 5, tenant-graph grounding 10 |
| Microsoft Foundry Agent Service | Pay for underlying resources | No separate runtime fee for prompt and workflow agents; models, tools, storage, hosted compute and memory billed separately (unit rates not shown on the page we retrieved) |
| Google Gemini Enterprise | Seat editions plus consumption | Pay-as-you-go edition: no license fee, $5 per GiB per month for storage and indexing, usage at Agent Platform rates; Standard, Plus and Frontline seats via sales |
| Amazon Bedrock AgentCore | Per-component usage, no minimum | Runtime $0.0895 per vCPU-hour plus $0.00945 per GB-hour (v1); Gateway $0.005 per 1,000 API calls; Web Search $7 per 1,000; Policy $0.000025 per request; Observability via CloudWatch |
| ServiceNow AI Agents + AI Control Tower | Subscription tiers plus metering | Not published; Foundation, Advanced and Prime tiers; contact sales |
| IBM watsonx Orchestrate | Monthly tiers | 30-day trial; Essentials from $578/month; Standard from $6,360/month; Premium custom |
| Oracle AI Agent Studio / OCI | Included with Fusion; OCI consumption | AI Agent Studio at no additional cost for Fusion customers (2025 announcement); OCI Generative AI Agents $0.0030 per 10,000 transactions; web search $10 per 1,000 requests |
| Databricks Agent Bricks | Usage of underlying Databricks services | No single price; billed at existing product rates, tracked per second |
| UiPath Agents + Maestro | Platform Units or Agent Units | 0.16 / 0.2 / 0.4 Platform Units per hosted LLM call (basic/standard/premium); Maestro Orchestrate 1 unit per instance; no public dollar value per unit |
| OpenAI Frontier / AgentKit | Sales-led enterprise; API usage | Frontier via OpenAI sales; AgentKit at standard API rates; ChatKit storage $0.10 per GB-day after 1 GB free; ChatGPT Business standard seats $20 (annual) or $25 (monthly) |
| Automation Anywhere | Sales-led | Not published; Community Edition free with limits |
| Kore.ai Artemis | Sales-led | Not published; also sold through AWS Marketplace |
| Glean | Seats plus Model Hub commitment or FlexCredits | No public dollar prices; per-run FlexCredit use is about 7 at the median and 114 at the 90th percentile |
What drives the real cost
- Actions, not chats. Salesforce, Microsoft and UiPath charge per agent action or LLM call, so one multi-step task can incur several charges.
- Model tokens. Developer platforms (AgentCore, Foundry, Agent Bricks, OCI, OpenAI API) bill model inference separately from the agent runtime.
- Retrieval, memory and storage. Vector storage, indexing and memory are separate line items at AWS, Oracle, Google and Microsoft.
- Observability and evaluation. AWS bills observability through CloudWatch and charges for evaluations. Databricks bills judge requests.
- Included allowances. Some licenses, such as Microsoft 365 Copilot for qualifying employee use and the Salesforce editions with bundled credits, offset metered costs.
How to estimate a budget
- Pick 2–3 target workflows and estimate monthly runs.
- Estimate the steps per run: tool calls, retrievals and model calls.
- Multiply by each vendor's published unit rates, and add storage and observability.
- For quote-only vendors, ask for the same scenario in writing, including overage rates.
Sources: each vendor's official pricing and licensing pages, listed on our methodology page. We have not run billed workloads ourselves.