Salesforce Launches AI Agents Marketplace

DIY AI 'clearly is not working for lots of people,' Salesforce EVP Brian Landsman tells CRN.

Salesforce has launched a new marketplace for its Agentforce AI platform that should allow partners to build and monetize agentic artificial intelligence components.

The marketplace, AgentExchange, is billed as accelerating AI agent deployment with a library of templates and add-ons meant to save users time and resources, according to the San Francisco-based enterprise applications vendor.

The marketplace launches with more than 200 initial partners and hundreds of actions. Salesforce announced the marketplace ahead of its TDX developer conference, which runs through Thursday in San Francisco.

Brian Landsman, Salesforce’s executive vice president and general manager of global business development and partnerships, told CRN in an interview that AgentExchange builds on the foundation of Salesforce’s AppExchange marketplace–now about 20 years old with more than 7,000 partners, 9,000 listings and 13 million app installs to date.

“We want to focus on the ability to bring third-party offerings direct ro users to be able to build these agents on a platform that does a lot of the work that you would otherwise have to go DIY, (do it yourself), which clearly is not working for lots of people,” said Landsman, referring to a common criticism Salesforce has lobbed at rival AI vendors, notably Microsoft.

“It's not working for lots of customers. It’s certainly not working for lots of partners,” he added.

[RELATED: Salesforce Q4 2025 Earnings: CEO Benioff Dismisses Threat To SaaS From Agentic AI]

Salesforce AgentExchange

Salesforce and Microsoft executives have battled publicly over which vendor brings the better AI offer for customers, with Salesforce CEO and co-founder Marc Benioff dismissing wares from Microsoft and other rivals as “false agents” and Microsoft CEO Satya Nadella suggesting that humans may not need to interact with software-as-a-service (SaaS) platforms such as Salesforce in the agentic era.

Allen Mann, director of the AI Innovation Hub at Salesforce partner Slalom–No. 27 on CRN’s 2024 Solution Provider 500–told CRN in an interview that his company is one of the few global system integrators chosen to be part of Salesforce’s Agentforce Partner Network.

Slalom has been building up its agentic AI expertise for about 24 months, creating “a really meaningful and impactful way for Slalom to show that expertise as industry-focused consultants that can help bring real, pragmatic, agentic solutions to our clients that institutionalize the needs that they have,” Mann said.

Slalom has been doing events for customers to learn more about how to use agents to transform their organizations, talking to hundreds of customers a week. Mann hired up his team to number about 25 people to meet new demand.

“Us being able to build things as a trusted partner in that way and show up to our clients with our level of industry expertise is key,” Mann said.

The AgentExchange helps bring enterprise-level trust to the agent era, Alice Steinglass, Salesforce’s executive vice president and general manager for developer experience and AppExchange, told CRN in an interview.

AI large language models (LLMs) “are fun–everybody can play with them,” Steinglass said. “The challenge is, how do I take that LLM and connect it to my data in a way that's secure and govern, where I have the right controls and permissions, and when that agent is using the data, they can access all the right data and only the right data. And I don't want to set that up with 5,000 agents separately.”

All actions have passed a security review are backed by customer reviews, according to Salesforce.

AgentExchange allows partners to sell four types of agentic building blocks. Actions are the jobs Agentforce agents can do, with flows and prompts allowing for industry-specific component tailoring. Topics focus and refine agent behavior, grouping actions and instruction around one task or job so that the agent follows guardrails and delivers consistent results.

Another building block is prompt templates, which leverage pre-written, reusable prompts for consistent interactions. These templates help agents gather information, assist users and achieve specific goals.

The fourth building block, agent templates, combines multiple topics and leverages partner-created actions to deliver an AI solution, with metadata and global instructions spanning topics, according to Salesforce.