Agentforce Studio
Agentforce Studio is the dedicated Setup-area workspace inside Salesforce for designing, configuring, testing, and deploying Agentforce Agents.
Definition
Agentforce Studio is the dedicated Setup-area workspace inside Salesforce for designing, configuring, testing, and deploying Agentforce Agents. It bundles Agent Builder (the agent configuration UI), Prompt Builder (the prompt template editor), Model Builder (where admins configure which LLM each agent uses), the Agentforce Testing Center (synthetic conversation testing), and an agent monitoring dashboard. Studio is the umbrella experience that ties every Agentforce configuration tool together so admins do not have to jump between disjoint Setup menus.
Studio is where the entire agent lifecycle happens. Admins start in Agent Builder to define Topics and Actions, drop into Prompt Builder to author prompt templates, use Model Builder to pick the LLM (Atlas Reasoning Engine, GPT, or another supported model), test in the Testing Center, and monitor production performance from the dashboard. The integrated workspace reduces the friction of building agents compared to the early Agentforce releases where each tool lived in a separate Setup area. It is the workbench, not a runtime product.
How Agentforce Studio brings the agent lifecycle into one Setup workspace
Agent Builder: the configuration core
Agent Builder is the primary UI inside Studio. Each Agent is a configured combination of Topics, Actions, Instructions, Channels, and Grounding. Admins drag Actions onto Topics, write Instructions in natural language, configure which LLM the Agent uses, and pick deployment channels. The Builder also includes a Test panel for sandbox-mode conversations during development, so admins can iterate without leaving the workspace.
Prompt Builder: prompt template authoring
Prompt Builder is the editor for Prompt Templates: parameterized natural-language prompts that produce structured outputs. Templates include input variables, instructions, grounding references, and output formatting hints. Prompt Templates can be invoked from agents as actions, from flows as actions, or directly via the REST API. Studio integrates Prompt Builder so admins authoring agents can build new prompt templates without leaving the workspace.
Model Builder: choosing the LLM
Model Builder is where the underlying language model is configured per agent or per use case. Salesforce ships Atlas Reasoning Engine as the default, with optional access to GPT (via Salesforce-managed OpenAI integration), Anthropic Claude, and other supported models. Each model has different cost, latency, and capability characteristics; Model Builder lets admins switch without rebuilding agents. The Einstein Trust Layer applies consistently across models, so safety guardrails persist.
Agentforce Testing Center: synthetic conversation tests
Testing Center runs synthetic conversations against draft agents and scores responses against expected outcomes. Tests cover happy path, edge cases, and adversarial cases. Admins build test suites per Topic; each test specifies an input message, expected Action calls, and expected response content. The Testing Center reports pass rates, failure reasons, and regression diffs between agent versions. This is the framework that makes agents shippable to production at quality.
Monitoring and the production dashboard
Studio includes a monitoring dashboard for deployed agents: conversation count, escalation rate, Action call distribution, average response time, customer satisfaction signals where the channel surfaces them. The dashboard is the operations-team view of agent performance, separate from the agent-build experience. Trends spotted here (escalation rate climbing, certain Topic resolution dropping) drive iteration back to Agent Builder.
Channel deployment and version management
From Studio, admins publish agents to channels: Service Console (internal), Experience Cloud (customer), Slack, WhatsApp, voice. Each channel has its own configuration: authentication, styling, message limits. Agent Versions are managed within Studio; activating a new version updates the channel runtime within minutes. Rollback to a prior version is a one-click operation if the new version misbehaves.
How Studio relates to the Agentforce platform
Studio is the build-and-test workspace; Agentforce is the runtime platform that hosts deployed agents. Studio configurations deploy to Agentforce, which then runs the agents 24x7, handles channel routing, applies the Einstein Trust Layer, and logs conversations for monitoring. Customers interact with Agentforce-hosted agents, not with Studio directly. Studio is admin-facing only; the deployed agents are the customer-facing layer.
Using Agentforce Studio to build, test, and deploy an agent
Studio brings the full agent lifecycle into one workspace. The pattern: open Studio, build the agent in Agent Builder, author needed Prompt Templates in Prompt Builder, pick a model in Model Builder, test in Testing Center, publish to a channel.
- Open Agentforce Studio
Setup, Quick Find, Agentforce Studio. The workspace lists every agent in the org plus links to Agent Builder, Prompt Builder, Model Builder, and Testing Center. Pick an existing agent or create a new one.
- Build or refine the agent in Agent Builder
Define Topics, attach Actions, write Instructions. Use the inline Test panel to validate behaviour as you build. Save the draft version frequently; Studio supports versioned drafts so you can roll back if needed.
- Author Prompt Templates as needed
For content-generation Actions, open Prompt Builder. Author the prompt with input variables, instructions, grounding references. Test the template in Prompt Builder before attaching to the agent.
- Configure the LLM in Model Builder
Pick the model per agent or per Topic. Atlas Reasoning Engine is the default; switch to GPT or another supported model if the use case warrants. Trust Layer applies consistently across models.
- Test in the Agentforce Testing Center
Build a test suite per Topic. Run the suite. Review failures and iterate on Agent Builder until pass rates are acceptable. Activate and deploy to a channel.
Agent Builder, Prompt Builder, Model Builder, Testing Center, monitoring dashboard. Pick the tool per task within the integrated workspace.
Atlas Reasoning Engine (default), GPT, Anthropic Claude, other supported models. Trade cost, latency, capability per use case.
Run the test suite before every deployment. Continuous testing (every save) catches regressions earlier than gate-only testing.
Studio manages versioned drafts per agent. Activate, rollback, or compare versions from the workspace.
- Studio is the build workspace, not the runtime. Agents run on the Agentforce platform after deployment; Studio is admin-facing only.
- Switching the underlying LLM mid-development can change agent behaviour subtly. Re-test after model switches before assuming behaviour is preserved.
- Testing Center coverage matters. Agents that pass shallow tests still fail in production; build deep test suites including adversarial cases.
- Channel deployment is per-version. Activating a new agent version in Studio updates the channel runtime; coordinate with operations before changing production agents.
- Studio updates frequently with Agentforce releases. Documentation can lag the UI; reference the current Salesforce Help when in doubt.
Trust & references
Cross-checked against the following references.
- Agentforce Studio OverviewSalesforce Help
- Agent Builder OverviewSalesforce Help
Straight from the source - Salesforce's reference material on Agentforce Studio.
- Prompt Builder OverviewSalesforce Help
- Agentforce Testing CenterSalesforce Help
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What is Agentforce Studio?
Q2. Which of these would you find in Agentforce Studio?
Q3. Why is Agentforce Studio valuable in large organizations?
Discussion
Loading discussion…