Agent builders, in the vendors’ own words
Every cell quotes the vendor. Checked 18 Nov 2026: 68 of 68 quotes still on their pages. How
Agent builders, feature by feature
- Yes
- Partly
- No
- Not documented
- Words in cells are each vendor’s own name for the feature. Click a cell for its source.
Can it search or read company apps such as SharePoint, Google Drive or Slack? Power Platform connectors let agents use prebuilt tools and triggers to reach other apps and services.
By using connectors, you can connect your accounts and use prebuilt tools and triggers to build your apps and workflows.
Can it search or read company apps such as SharePoint, Google Drive or Slack? Agents connect to Salesforce data, Flows, Apex, MuleSoft APIs and external systems.
You can connect Agentforce agents to the Salesforce tools and data you already use — like Flows, Apex, MuleSoft APIs, CRM data, and knowledge articles.
Can it search or read company apps such as SharePoint, Google Drive or Slack? Agents act across SAP and non-SAP systems.
Build and extend agents that reason through tasks and take action across SAP and non-SAP systems.
Can it search or read company apps such as SharePoint, Google Drive or Slack? Agents use APIs, knowledge bases and flows as tools; no list of app connectors is on this page.
Equip your agent with tools by integrating APIs for actions and knowledge bases.
Can it search or read company apps such as SharePoint, Google Drive or Slack? Workday exposes its own data to agents; reading other company apps is not described here.
Connect Workday to any agent. Agent-Ready Tools are engineered for agents with the structure and context agents need to act accurately, every time.
Can it connect to tools through the Model Context Protocol (MCP)? Agents can connect to MCP servers; tools and resources are supported, prompts are not.
Copilot Studio currently supports MCP tools and resources.
Can it connect to tools through the Model Context Protocol (MCP)? Salesforce describes MCP support with governance; its page uses future tense, so check current availability.
Agentforce will provide secure, enterprise-grade agentic interoperability by leveraging Model Context Protocol (MCP).
Can it connect to tools through the Model Context Protocol (MCP)? Custom agents support MCP and A2A; SAP said GA was planned for December 2025.
New capabilities include AI-assisted agent design, system-triggered agents, extensibility of SAP-delivered Joule Agents, centralized enterprise-grade agent monitoring, support for Agent-to-Agent (A2A) protocol, and support for Model Context Protocol (MCP).
Can it connect to tools through the Model Context Protocol (MCP)? Agents can call external MCP servers through the MCP client.
ServiceNow AI Agents can reach out to external tools, data, and systems via the Model Context Protocol (MCP) client.
Can it connect to tools through the Model Context Protocol (MCP)? Workday connects to agents over open protocols including MCP and A2A.
This allows customers to connect through Teams, Gemini, and open protocols like MCP and A2A.
Can people or builders choose models from more than one AI company? Builders can pick Anthropic, xAI or Mistral models as well as OpenAI GPT; Claude may process data cross-region.
You can also add external AI models from Anthropic, xAI, or Mistral to your agent.
Can people or builders choose models from more than one AI company? Admins set an org default model and can override it per agent with OpenAI, Anthropic or Google models.
We recommend that you use GPT 4.1 ( sfdc_ai__DefaultGPT41 ), Claude Haiku 4.5 ( sfdc_ai__DefaultBedrockAnthropicClaude45Haiku ), or Gemini 3.5 Flash ( sfdc_ai__DefaultVertexAIGemini35Flash ) because we’ve thoroughly tested these models with agents.
Can people or builders choose models from more than one AI company? Joule Studio builders pick from 70+ models; SAP's own tutorials show OpenAI and Anthropic models.
With Joule Studio, business users and developers can use no-code and pro-code tools to build and extend agents, choosing from more than 70 AI models.
Can people or builders choose models from more than one AI company? OpenAI GPT is the default; admins can switch agents to ServiceNow's own model, Anthropic Claude or Google Gemini.
The orchestrator LLM can also be configured to use other models such as NowLLM, Claude, and Gemini.
Can people or builders choose models from more than one AI company? Agents on Workday's Agent Harness can run on models from six companies, including Anthropic, Google and OpenAI.
Run agents on foundation models from Anthropic, Google, and OpenAI, or open models from NVIDIA, Meta, and Mistral.
Can a business user build an agent without writing code? A graphical low-code studio; no extensive technical background needed.
Microsoft Copilot Studio is a graphical, low-code studio for building and managing AI-powered agents and workflows.
Can a business user build an agent without writing code? Business users build agents in plain language on a low-code canvas; the builder is included with Agentforce.
Every team can build reliable, powerful agents with a low-code AI agent builder.
Can a business user build an agent without writing code? Business users build agents through a guided low-code experience; developers can go pro-code.
Business users applying their process and domain expertise to create AI agents and workflows through a guided, low-code experience.
Can a business user build an agent without writing code? Build and customise agents with a natural language interface and guardrails.
You can easily create specialized AI agents, set guardrails, and automate tasks using a natural language interface.
Can a business user build an agent without writing code? Build custom agents in plain language, with governance built in.
Empowers your team to build custom agents in plain language, codes using your preferred tool, and ensures every agent deployed has built-in governance.
Can one agent hand a task to another agent, including agents built elsewhere (for example over A2A)? Agents can hand work to child agents, other Copilot Studio agents, or outside agents over A2A.
You can connect your custom agent to another agent that supports the A2A protocol.
Can one agent hand a task to another agent, including agents built elsewhere (for example over A2A)? Agents delegate tasks to registered third-party agents over A2A; outside agents can also call Agentforce.
Agentforce A2A Outbound is the native way for Agentforce agents to collaborate directly with a third party agent.
Can one agent hand a task to another agent, including agents built elsewhere (for example over A2A)? Joule hands tasks to custom and third-party agents over A2A; SAP says its two-way agent gateway is not yet GA.
Pro-code agents, which run in their own independent runtime environments, are integrated with Joule using the open Agent2Agent (A2A) protocol
Can one agent hand a task to another agent, including agents built elsewhere (for example over A2A)? The orchestrator runs teams of agents, and A2A links them to agents from other vendors.
Embedded within Action Fabric , the Agent2Agent (A2A) protocol enables communication among ServiceNow and third-party AI agents.
Can one agent hand a task to another agent, including agents built elsewhere (for example over A2A)? Workday agents and outside agents, such as Gemini Enterprise ones, can hand tasks to each other over A2A.
The partnership supports Agent-to-Agent (A2A), Agent-to-UI (A2UI) and Model Context Protocol (MCP) approaches so AI agents can share information and autonomously hand off tasks to each other in real time, all within a single workflow.
Can an agent start by itself when an event happens or on a schedule? Agents can start on connector events or a recurrence schedule; each trigger run uses Copilot Credits.
Unlike topic triggers, which require input from a user, event triggers allow an agent to act autonomously in response to the specified event.
Can an agent start by itself when an event happens or on a schedule? Any active agent can be started from a flow or Apex, so a record change or schedule can launch it.
Now, invocable agent actions allow you to trigger any active Agentforce agent directly from Apex and flows.
Can an agent start by itself when an event happens or on a schedule? Joule Agents can run in the background, started by a business event, a schedule or a system signal.
In system-triggered mode, agents run in the background, initiated by a business event, a schedule, or a system signal rather than a user request.
Can an agent start by itself when an event happens or on a schedule? Agents can start when a record changes or a set condition is met, with no user prompt.
AI agent builders can create triggers when creating an AI Agent or Agentic workflow that calls the agent when a condition or objective is observed.
Can an agent start by itself when an event happens or on a schedule? Scheduled and event-started custom agents need the Sana Enterprise SKU; autopilot Workday agents are early adopter only.
Users can trigger agents manually, on a schedule, or through third-party events.
Can an agent stop and wait for a person to approve a step before it acts? Agent flows can pause for a person's approval or input; multistage and AI approvals are still in preview.
Human in the loop: Actions that require human intervention, such as approval requests or providing information.
Can an agent stop and wait for a person to approve a step before it acts? An action can be set so the person in the conversation must confirm before the agent runs it.
Indicates whether the customer must confirm before the agent runs the action.
Can an agent stop and wait for a person to approve a step before it acts? Joule Agents need a person's explicit approval for high-risk steps.
High-risk processes always require explicit approval and all agent activity is logged and traceable.
Can an agent stop and wait for a person to approve a step before it acts? People can review an agent's plan and approve it before it runs.
For example, human service agents can review an AI Agent-generated plan and provide feedback or explicit approval before execution or changes.
Can an agent stop and wait for a person to approve a step before it acts? Sana asks a person to confirm actions that need approval; pending approvals wait in the Sana Inbox.
Prompts users to review and confirm actions that require human-in-the-loop approval before proceeding.
Can builders run the agent against test cases or simulated conversations before it goes live? Builders run test sets of single questions or full conversations and score the answers before release.
Agent evaluation lets you generate tests that simulate real-world scenarios for your agent.
Can builders run the agent against test cases or simulated conversations before it goes live? Builders generate test cases and run simulated conversations at scale before an agent goes live.
Testing Center simulates behavior at scale with state injection and AI-driven evals — so you can pressure-test your agents before going live.
Can builders run the agent against test cases or simulated conversations before it goes live? Joule Studio generates and runs automated tests for a new agent before it is deployed.
It runs an automated validation suite derived from the specification - you do not need to initiate it manually.
Can builders run the agent against test cases or simulated conversations before it goes live? Builders run evaluations of multi-step agent tasks against sample records before rollout.
AI Agent evaluations allow you to test complex, multi-step tasks and orchestration logic before rollout.
Can builders run the agent against test cases or simulated conversations before it goes live? Agent Harness evals measure an agent's accuracy before deployment; generally available since October 2026.
Measure accuracy before deployment and give auditors a step-by-step trace of what the agent did.
Can admins export or query a log of who did what? Admins use built-in auditing in Microsoft Purview to filter agent events and activities.
As an administrator, you can use the built-in auditing capabilities of Copilot Studio to secure and govern your environments.
Can admins export or query a log of who did what? Audit Trail tracks AI agent actions and outputs for governance review.
For example, the Audit Trail feature provides the data you need to track AI agent actions and outputs, ensuring AI usage complies with your organization’s security, privacy, regulatory, and AI governance policies.
Can admins export or query a log of who did what? Audit logging is stated for SAP AI Core, the AI platform layer, not specifically for Joule agents.
Comprehensive audit logging to protect customer data at every stage of processing.
Can admins export or query a log of who did what? Control Tower offers auditable compliance gates for AI; a plain export of agent logs is not described.
Super Intelligence Control Tower provides you with flexible, automated, auditable, and traceable compliance gates.
Can admins export or query a log of who did what? Admins can view a record of everything agents do.
Quickly view a record of everything agents do, and prove the agents are actually working.
Can the company choose the region where its data is stored? Organizations choose the Azure geography where data is stored; some generative AI can cross regions.
Microsoft Copilot Studio allows organizations to choose where their data is stored, providing flexibility to meet regional data residency requirements.
Can the company choose the region where its data is stored? In-country storage is offered on Hyperforce in listed countries, so it depends on region and product.
Hyperforce lets businesses, particularly those operating regionally, store data within country borders.
Can the company choose the region where its data is stored? Searched sap.com AI Core, Joule Studio and pricing pages and help.sap.com results; no region-choice statement for Joule agents found.
No statement from SAP on its own pages. Searched sap.com AI Core, Joule Studio and pricing pages and help.sap.com results; no region-choice statement for Joule agents found.
Can the company choose the region where its data is stored? ServiceNow says AI processing stays in the customer region, with routing policies set in Control Tower.
AI data processing is confined to the customer’s region, adhering to local regulations such as GDPR and supporting jurisdiction-specific compliance requirements.
Can the company choose the region where its data is stored? Searched Workday trust, privacy, compliance and AI pages; no statement letting customers choose a data region for agents.
No statement from Workday on its own pages. Searched Workday trust, privacy, compliance and AI pages; no statement letting customers choose a data region for agents.
Can the company encrypt its data with keys it controls (customer-managed keys)? Uses Power Platform customer-managed keys; works only in Managed Environments, older data keeps Microsoft keys.
Copilot Studio supports CMK, which lets customers control access to their data within Copilot Studio.
Can the company encrypt its data with keys it controls (customer-managed keys)? Customer-managed keys come with Shield, a paid add-on for the whole platform; nothing agent-specific.
Salesforce Shield is a paid add-on bundle that includes four natively integrated tools: Platform Encryption (AES-256 field- and file-level encryption at rest with customer-managed keys)
Can the company encrypt its data with keys it controls (customer-managed keys)? Searched Joule Studio, SAP AI Core and SAP Key Management Service pages; BYOK exists for SAP cloud services but Joule is not named.
No statement from SAP on its own pages. Searched Joule Studio, SAP AI Core and SAP Key Management Service pages; BYOK exists for SAP cloud services but Joule is not named.
Can the company encrypt its data with keys it controls (customer-managed keys)? Bring-your-own-key is a platform-wide Cloud Encryption option; ServiceNow says nothing specific to AI Agents.
Gain the flexibility to use either the ServiceNow Managed Key or "Bring Your Own Key" option.
Can the company encrypt its data with keys it controls (customer-managed keys)? BYOK covers Workday platform data at rest; Workday says nothing specific about agent data or model calls.
Additionally, customers have the option to implement bring-your-own-key capability to retain full control of their root encryption keys.
Is business data excluded from model training by default? Customer data is not used to train Microsoft models unless the tenant admin opts in.
Are NOT used to train or improve Microsoft AI models, unless your tenant admin opts in to sharing data with us.
Is business data excluded from model training by default? Prompts and responses are not stored or used to train the third-party LLMs. Part of the Einstein Trust Layer.
Zero data retention is a strict policy where the prompts and generated responses are never stored or used to train the underlying third-party large language models
Is business data excluded from model training by default? SAP says customer data is not used to train SAP models or by third-party LLM vendors.
Customer data is never used to train SAP models or shared with third-party vendors for training LLMs.
Is business data excluded from model training by default? Third-party model data is not used for training; sharing for ServiceNow model improvement is stated separately.
For integrated third-party models, we ensure that your data remains within our secure platform and is not used to train external models.
Is business data excluded from model training by default? Workday never shares customer data to train third-party public models; some ML features use opted-in tenant data.
We never share customer data to train third-party public models.
Does the vendor state SOC 2 Type II and ISO 27001 for this product? Page states SOC audit and ISO 27001; it does not say SOC 2 Type II.
Microsoft Copilot Studio has been audited to be compliant with SOC.
Does the vendor state SOC 2 Type II and ISO 27001 for this product? Salesforce lists a SOC 2 report and an ISO 27001:2022 certificate for Agentforce and its AI platform.
SOC 2 Report - AI Platform and Agentforce
Does the vendor state SOC 2 Type II and ISO 27001 for this product? Stated for SAP AI Core, the AI layer; ISO 27001 is also claimed on that page, though misspelled.
Certification for SOC1 and SOC2.
Does the vendor state SOC 2 Type II and ISO 27001 for this product? No SOC 2 or ISO 27001 statement found for AI Agents on servicenow.com pages searched (product, Control Tower, community, trust paths).
No statement from ServiceNow on its own pages. No SOC 2 or ISO 27001 statement found for AI Agents on servicenow.com pages searched (product, Control Tower, community, trust paths).
Does the vendor state SOC 2 Type II and ISO 27001 for this product? ISO 27001 scope includes Sana Agents; Workday also publishes a SOC 2 Type II report.
Information Security Scope (ISO 27001, 27017, 27018): Adaptive Planning, Analytics and Reporting, Contract Lifecycle Management (CLM) and Contract & Document Intelligence, Financial Management, HiredScore, Human Capital Management, Payroll, Peakon, Platform and Product Extensions, Spend Management, Student, Talent Management, Workforce Management, and Sana (Core/Enterprise, Learn, Agents).
Will the vendor sign a HIPAA business associate agreement that covers this product? Copilot Studio is in scope of Microsoft's HIPAA business associate agreement.
Microsoft Copilot Studio is covered under the Health Insurance Portability and Accountability Act (HIPAA) Business Associate Agreement (BAA).
Will the vendor sign a HIPAA business associate agreement that covers this product? The BAA lists named agents (Service Agent, Assistant, SDR, Voice), not every agent, and requires Data Masking.
With respect to Agentforce Services and Prompt Builder and to the extent within Customer’s control, Customer must enable, maintain, and use Data Masking in the Einstein Trust Layer configuration for Customer’s HIPAA Covered Services.
Will the vendor sign a HIPAA business associate agreement that covers this product? Searched the SAP Trust Center compliance page and Joule, Joule Studio and AI Core pages; no HIPAA or BAA statement.
No statement from SAP on its own pages. Searched the SAP Trust Center compliance page and Joule, Joule Studio and AI Core pages; no HIPAA or BAA statement.
Will the vendor sign a HIPAA business associate agreement that covers this product? Searched ServiceNow trust and compliance pages and the Now Assist privacy and model FAQs; no BAA statement for AI Agents.
No statement from ServiceNow on its own pages. Searched ServiceNow trust and compliance pages and the Now Assist privacy and model FAQs; no BAA statement for AI Agents.
Will the vendor sign a HIPAA business associate agreement that covers this product? A third-party HIPAA attestation for Workday Enterprise Products; no BAA or agent-specific statement found.
Workday has completed a Health Insurance Portability and Accountability Act (HIPAA) third-party attestation for the Workday Enterprise Products
Is this product FedRAMP authorized for US government use? Applies to the US Government plans (GCC and GCC High), designed for the FedRAMP High level.
Copilot Studio is authorized as a service within the Azure Government FedRAMP ATO.
Is this product FedRAMP authorized for US government use? Agentforce is FedRAMP High authorized for US public sector use in Salesforce Government Cloud.
Salesforce products including Agentforce, Data Cloud, Marketing Cloud, and Tableau Next have achieved FedRAMP High authorization
Is this product FedRAMP authorized for US government use? SAP Trust Center links FedRAMP authorizations for some SAP services; no page says Joule or Joule Studio is authorized.
No statement from SAP on its own pages. SAP Trust Center links FedRAMP authorizations for some SAP services; no page says Joule or Joule Studio is authorized.
Is this product FedRAMP authorized for US government use? AI Agents, Orchestrator and AI Agent Studio are GA at FedRAMP High in ServiceNow's Government Community Cloud.
general availability at the FedRAMP High Authorization level of ServiceNow AI Agents, AI Agent Orchestrator, and AI Agent Studio for our Government Community Cloud (GCC) customers in the U.S.
Is this product FedRAMP authorized for US government use? Workday Government Cloud is FedRAMP Moderate; Workday does not say its AI agents run in that boundary.
Workday is FedRAMP Authorized status at the Moderate security impact level for Workday Government Cloud.
Latest change, 28 Oct 2026: Microsoft Copilot Studio. Microsoft Copilot harness, used by Copilot Studio, opens via Foundry. All changes
Who sells what
A few companies sell in most of these categories, so start from the row you already pay for. Everyone else sells one kind.