ChatGPT
Claude
PerplexityAI Tools for Salesforce in 2026: Current Names and What Each Does
Most articles answering this question list tools that no longer carry those names. Einstein Copilot, Einstein GPT, Einstein 1 Studio and Data Cloud have all been renamed or absorbed, and searching for them now returns documentation for something else. The current native stack is Agentforce for agents, Agentforce Builder for building them, Agentforce Vibes for development, Prompt Builder for prompts, Data 360 for unified data, and the Einstein Trust Layer underneath all of it.
Start with the map, because a name mismatch is the single biggest source of confusion on this topic.
Einstein-era names were replaced between 2024 and 2026
Name you may be searching for | What it is now | Status |
|---|---|---|
Einstein Copilot | Agentforce | |
Data Cloud | Data 360 | Renamed, same product |
Agent topics | Subagents | Renamed April 2026, no functional change |
Einstein for Developers | Agentforce Vibes | Superseded; the earlier dev assistant is deprecated |
Einstein Bots | Agentforce Service Agent | Superseded for new builds |
Base, Ridge, Crest, Summit partner tiers | Summit and Select | Replaced 2 March 2026 |
Two of these matter operationally rather than cosmetically. The topics to subagents rename means documentation and vendor conversations now mix both terms freely. And the partner tier change means any article telling you to look for a Crest partner is describing a filter that no longer exists.
Salesforce native AI tools group by who uses them
The native stack is easier to hold in your head by role than by product family.
For administrators. Agentforce Builder for configuring agents without code, reaching general availability in Spring ’26 alongside a legacy builder Salesforce still documents separately. Prompt Builder for reusable grounded prompts. Setup with Agentforce for creating objects, troubleshooting user access and answering configuration questions inside Setup itself. Einstein Prediction Builder and Next Best Action remain for predictive scoring and recommendation logic.
For developers. Agentforce Vibes, an org-aware development environment that reads your metadata and schema and keeps code, schema and metadata inside the Agentforce Trust Boundary. Agent Script, an agent definition language Salesforce open sourced at TDX 2026, which lets you declare where an agent reasons and where it must follow deterministic logic. Salesforce Code Analyzer for static analysis.
For architects and data teams. Data 360 for unification and grounding, with free ingestion and consumption-priced actions. The Einstein Trust Layer for zero data retention, dynamic grounding, prompt defence, toxicity detection and the audit trail.
For quality and release. Agentforce Testing Center, which scores agent behaviour across coherence, completeness, conciseness, latency, instruction adherence and factuality rather than returning a simple pass.
Hosted MCP Servers opened the org to any AI client
The structural change of 2026 sits underneath all of the above. Salesforce Hosted MCP Servers reached general availability on 29 April 2026 for Enterprise Edition orgs and above.
A hosted MCP server is a Salesforce-managed endpoint exposing your org’s data, flows, Apex actions and queries to any AI client speaking the Model Context Protocol. Your existing permissions apply automatically, including CRUD, field-level security and sharing rules, and every transaction runs as the authenticated user rather than an anonymous service account.
That changes the native versus third-party question permanently. A general assistant connected through MCP can now read live org data under real permissions, which removes the historical argument that only native tools could see your data. What native tooling still holds is metadata awareness, since reading records is different from reading configuration.
Third-party AI tools for Salesforce fall into three groups
Group | What it does | Where it beats native tooling |
|---|---|---|
General assistants connected via MCP | Answers questions against live org data | Conversational range, and it works where your team already is |
Coding assistants | Writes Apex, Lightning Web Components, tests | Editor integration and multi-language work outside the platform |
Delivery platforms | Produces the project artifact chain grounded in org metadata | Analyst-phase artifacts that native tooling does not attempt |
The third group is where the native stack has the largest gap, because Salesforce builds tools for people configuring an org rather than for people running a project. Requirements, solution design documents, traceable user stories and test strategy sit outside what Agentforce Builder or Vibes set out to do. GetGenerative.ai operates in that group, with agents covering discovery through post-go-live support and output deploying into Salesforce, Jira and GitHub. The scoring criteria for that whole category are in platforms purpose-built for delivery.
Free Salesforce AI tiers cover more than teams assume
Several native capabilities carry no separate charge. Salesforce Foundations provides Agentforce Builder, Prompt Builder, Agent Script and Agentforce Vibes at no cost, along with 100,000 Flex Credits for Enterprise Edition orgs and above. Data 360 ingestion is free, with charges attaching to actions rather than to bringing data in. Digital Wallet, which tracks consumption in near real time, is provided at no charge for enabled products.
What does cost money is usage. Agentforce actions consume 20 Flex Credits each, which is $0.10, and Data 360 actions draw on the same credit pool. The practical implication is that experimenting is close to free and scaling is metered, so the number worth modelling early is actions per conversation rather than licence count.
Choosing between native and third-party Salesforce AI tooling
Four questions settle most of it.
- Does the task need metadata or records? Records favour anything connected through MCP. Metadata favours native tooling or a platform that analyses configuration directly.
- Does the output need to deploy? If yes, you need something that writes to the org rather than returning text.
- Where must the data stay? Native tooling keeps metadata inside the Salesforce trust boundary by design. Any third-party tool needs that answer in writing.
- Is the work configuration or project delivery? Salesforce builds for the first. The second is where third-party platforms earn their place.
Most teams end up running both, and that is a reasonable outcome rather than indecision. The distinction between a general assistant and a purpose-built agent is worked through in generic chat versus real agents.
Where teams waste effort on Salesforce AI tooling
Mistake | What happens | Correction |
|---|---|---|
Searching for retired product names | You read documentation for something else | Use the name map above before researching |
Assuming MCP access equals metadata awareness | Design advice conflicts with the org | Test a structural question, not a data question |
Pasting org exports into a consumer chat tool | Data leaves the boundary with no audit trail | Connect under the org’s permission model instead |
Modelling cost by licence count | Consumption arrives unbudgeted | Model actions per conversation |
Buying a third-party tool for a native capability | Paying twice for prompt building or agent config | Check Foundations first |
Treating agent testing as conventional testing | Sign-off has no threshold | Score against named quality dimensions |
Salesforce’s own State of IT research found nearly 40 percent of new applications now include AI features, which is why the tooling question moved from optional to structural inside a single release cycle.
The current Salesforce AI tool stack, condensed
Need | Current tool |
|---|---|
Build an agent without code | Agentforce Builder |
Define agent behaviour in code | Agent Script |
Develop with org awareness | Agentforce Vibes |
Reusable grounded prompts | Prompt Builder |
Unify and ground data | Data 360 |
Security and audit for generative AI | Einstein Trust Layer |
Test agents before production | Agentforce Testing Center |
Track consumption | Digital Wallet |
Connect an external AI client to the org | Salesforce Hosted MCP Servers |
Produce project artifacts | Third-party delivery platforms |
Agents settle the Salesforce AI tooling decision without a long evaluation
Native tooling covers configuration and agent work well, and covers project delivery barely at all. Teams delivering Salesforce projects rather than administering an org usually need something in the third group, and the fastest way to know is to run one real deliverable through a candidate and time how long you spend fixing the output. The Pro plan for consultants includes a trial period with full agent access for exactly that test.
Questions teams ask about AI tools for Salesforce
What are the AI tools in Salesforce?
Agentforce for agents, Agentforce Builder for configuring them, Agentforce Vibes for development, Agent Script for agent logic, Prompt Builder for prompts, Data 360 for unified data, Einstein Trust Layer for security and audit, and Agentforce Testing Center for validation. Prediction Builder and Next Best Action remain for predictive work.
What happened to Einstein?
The Einstein name largely gave way to Agentforce for agentic capability, with Einstein Copilot renamed to Agentforce and Data Cloud renamed to Data 360. The Einstein Trust Layer keeps its name and still underpins generative features across the platform.
Is Claude good with Salesforce?
General assistants including Claude can now connect to a Salesforce org through Hosted MCP Servers, reading live data under your existing permissions. They are strong on questions about records and weaker on questions about configuration, because reading records is not the same as reading metadata.
What AI agent does Salesforce use?
Agentforce is the agent platform, with Agentforce Service Agent for customer-facing use and employee-facing agents for internal work. Underneath, agents are composed of subagents and actions, grounded through Data 360 or knowledge sources, and governed by the Einstein Trust Layer.
Are there free AI tools for Salesforce?
Yes. Salesforce Foundations provides Agentforce Builder, Prompt Builder, Agent Script and Agentforce Vibes at no cost, plus 100,000 Flex Credits on Enterprise Edition and above. Data 360 ingestion is free, and Digital Wallet for tracking consumption is provided at no charge.
Can AI replace Salesforce CRM?
No. Agents operate on the CRM rather than in place of it, reading records, running actions and writing back under the platform’s permission model. What changes is the interface, since more work now happens through conversation than through page layouts.
