Skip to content
Webinar on 19 Nov: Customer meetings with AI agents Register now →
Article AI · Salesforce

Salesforce puts a price on AI agents: what Flex Credits mean for MCP and API calls

Tim Vogel September 23, 2026 · 2 min read

Salesforce is introducing a new billing model for agent-based access to its platform. Successful calls via the Model Context Protocol (MCP) and direct APIs by registered AI agents will consume Flex Credits. The change is part of Salesforce’s headless toolkit strategy and means you need to recalculate your AI architecture and budgets.

Agentic Identity separates AI from human users

Until now AI tools mostly ran under the identity of the human user. With the new “Agentic Identity” framework every AI agent gets its own credentials and permissions. As an administrator you control exactly which data an agent can access and what it may do. That lowers the risk of an external agent changing CRM data without authorisation because of overly broad rights.

The cost trap: Flex Credits for Headless Platform Interactions

Once an agent is registered, the Digital Wallet starts counting. Every successful call to Salesforce, whether via MCP or API, counts as a “Headless Platform Interaction” and costs Flex Credits. If you connect external models such as Claude or ChatGPT you pay twice: the model provider’s API costs and the Flex Credits at Salesforce.

External modelClaude, ChatGPTAPI costs
Agentic IdentityRegistered agentMCP / API
SalesforceDigital WalletFlex Credits
With an external model, costs arise in two places.

How much do Agentforce and Flex Credits cost?

Salesforce bills Agentforce by consumption. For external communication you usually pay a flat rate per conversation. Internal actions and MCP/API access cost Flex Credits. Every successful call by a registered agent draws down a defined quota.

What a Flex Credit costs and how high the multiplier for a Headless Platform Interaction is, Salesforce has not published yet. When metering starts there is a 30-day lead time. Only active production environments are billed. From summer 2025 Salesforce also plans per-user licences for employee-facing agents.

What are the use cases for AI agents?

Agents summarise cases in customer service or qualify leads in sales, for example. They access structured and unstructured CRM data and complete defined work steps on their own. Via integration platforms such as MuleSoft they can also connect to external systems such as ERP solutions.

How to prepare your Salesforce orgs for the new model

Classic API integrations without agents keep their existing pricing and security rules. For real AI agents you need to rebuild:

Migrate to OAuth credentials

MCP clients and API-based agents will need the OAuth credentials of the new Agentic Identity. Plan this effort into your developers' sprints early.

Use test environments

Sandboxes, scratch orgs and Developer Edition orgs are exempt from Flex Credit costs. Develop and test agent-based workflows there. Check the efficiency of your prompts in these environments before they consume real credits in production.

Secure data quality

Faulty data leads to unnecessary correction calls by the AI, and those cost credits. Clean up legacy data before you let agents loose on production data. Tools such as Data 360 (formerly Data Cloud) help with a clean structure.

Monitor the consumption of your Flex Credits in the Digital Wallet. Set up threshold alerts there to protect your budget from outliers caused by inefficient workflows.

Do you want to move existing AI integrations to Agentic Identity or are you planning your first Agentforce rollout? We check with you how to build your architecture so you avoid unnecessary Headless Platform Interactions. Get in touch for an architecture check.

Intro call

Want to know what would work in your company?

Thorsten Winkels, Managing Director Sales of Aviando, looks at your Salesforce landscape, your AI ideas and your marketing processes with you. Together you work out where to start and which steps your team can take on its own.

Thorsten Winkels Thorsten Winkels Managing Director Sales
Keep reading

More blog posts on AI and Salesforce

All posts
Marketing Cloud
Salesforce Marketing Cloud & AI agents: what the Dreamforce announcements mean for B2B
3 min read
Article Marketing

Salesforce Marketing Cloud & AI agents: what the Dreamforce announcements mean for B2B

AI agents like Piper and Hunter are changing B2B lead generation in Salesforce Marketing Cloud. Learn which data prerequisites the mid-market needs for them.

Read article Oct 15, 2026 · 3 min
Flex Credits
Flex Credits for MCP: why Salesforce's AI pricing endangers clean architectures
3 min read
Article Salesforce

Flex Credits for MCP: why Salesforce's AI pricing endangers clean architectures

Salesforce will soon charge Flex Credits for external AI agents via MCP. Learn why this can lead to data silos and how to protect your architecture.

Read article Oct 13, 2026 · 3 min
Agentforce
Salesforce Koa: what the new reasoning model for Agentforce means in day-to-day B2B work
3 min read
Article AI

Salesforce Koa: what the new reasoning model for Agentforce means in day-to-day B2B work

Salesforce Koa is an AI model specialised in CRM processes for Agentforce. Learn how it reduces errors and what that means for your AI strategy.

Read article Oct 1, 2026 · 3 min