Salesforce Dictionary - Free Salesforce GlossarySalesforce Dictionary
DictionaryAAgentforce Assets
AIAdvanced

Agentforce Assets

Agentforce Assets are the metadata building blocks used to define, configure, and deploy an Agentforce agent.

§ 01

Definition

Agentforce Assets are the metadata building blocks used to define, configure, and deploy an Agentforce agent. The asset family includes Agents (the top-level autonomous agent definition), Topics (the bundled groupings of intent and instruction that scope what the agent can handle), Actions (the discrete operations the agent can invoke, like creating a record or calling a flow), Prompt Templates (the structured instructions the agent uses for generative responses), and Permissions and Channels (where the agent runs and what data it can touch). Each asset is a Salesforce metadata record with its own object, lifecycle, and packaging behavior.

Asset-based composition is what distinguishes Agentforce from a one-off prompt or a single chatbot script. The same Action can be reused across Topics, the same Topic can be reused across Agents, and the same Prompt Template can ground multiple Actions. Reuse means an Agentforce deployment scales as a metadata project, not a copy-paste script collection. The trade is operational: agent assets must be governed, version-controlled, and deployed through Salesforce DevOps practices, not built informally in production.

§ 02

The asset types that make up an Agentforce agent

Agents: the top-level autonomous unit

An Agent is the named, configured autonomous unit the customer interacts with. It has an identity, a description of its purpose, a set of Topics it can handle, a Channel where it runs (sales console, service console, messaging surface, Slack), and a model selection for its reasoning. The Agent record is the package boundary; deploying an agent means deploying the Agent record plus all the Topics, Actions, and Prompt Templates it references. Customers usually have several Agents (a Sales Agent, a Service Agent, an Internal Knowledge Agent) rather than one all-purpose agent.

Topics: scoping what the agent can do

Topics are the next level of decomposition. A Topic represents a coherent area of capability (Process a refund, Schedule a service appointment, Answer product questions). Each Topic has a set of trigger utterances (the language patterns that route into the Topic), instructions (natural language guidance for how the agent should behave inside the Topic), and a list of Actions the agent can invoke. The Atlas Reasoning Engine selects a Topic by matching the incoming utterance against the Topics defined on the Agent.

Actions: what the agent can actually do

Actions are the discrete operations available to the agent. Standard Actions ship with Agentforce (search a knowledge article, create a case). Custom Actions wrap flows, Apex, or prompt templates as agent-callable units. Each Action has input parameters, an output schema, and a description the agent uses to decide when to call it. Actions are the integration boundary: they are what gives the agent the ability to take real effects in the world (move a record, send an email) versus only generating text.

Prompt Templates: the generative voice

Some Actions wrap Prompt Templates, which means the Action is itself a generative call. The template defines the system instruction, the grounding sources, the output schema, and the constraints. Prompt Templates are first-class assets, manageable in Prompt Builder, and reusable across Actions. The template-as-asset pattern means generative behavior is defined in one place and reused, rather than duplicated inside each calling Action.

Channels and where agents run

Channels define the surface where an Agent is reachable. Out of the box, Agentforce ships channels for the Sales console (Agentforce for Sales), Service console (Agentforce for Service), Slack (via the Slack integration), and Messaging for In-App and Web. Each channel has its own UX adapter that renders the agent's responses appropriately for the surface. The same Agent can run on multiple channels with channel-specific UX hints (richer cards in the console, plain text in Slack).

Permissions and the data boundary

Agentforce Agents respect Salesforce permissions like any other user. Each agent has an assigned user (or runs as the calling user) and inherits that user's profile, permission sets, and sharing rules. An agent cannot create a record the calling user does not have permission to create. This is critical for the security posture of an Agentforce deployment: the agent's authority is bounded by Salesforce's standard authorization model, not by an additional layer of agent-specific rules.

Deployment, packaging, and version control

Agentforce Assets are metadata. They can be deployed through change sets, unlocked packages, or Salesforce DX. Version control through Git is recommended for any production agent project. The asset family includes some metadata types that are recent and may have packaging caveats; check the current platform release notes before designing the deployment pipeline. The discipline that applies to flows and Apex classes applies to agents: pilot in a sandbox, version, deploy through a defined release path, do not edit in production.

§ 03

How to design an Agentforce agent as a composition of assets

Designing an Agentforce agent is a top-down asset composition: start with what the Agent should accomplish, decompose into Topics, decompose each Topic into Actions, wrap generative work in Prompt Templates.

  1. Define the Agent and its purpose

    Pick the audience (customer, employee, partner), the channel (Sales console, Service, Slack, messaging), and the assigned user identity. Write a clear purpose description; this drives Topic selection later.

  2. Decompose into Topics

    List the capability areas the Agent should handle. Each becomes a Topic. Keep Topics coherent: one Topic should represent a single capability area with clear boundaries, not a grab-bag.

  3. Identify or build the Actions per Topic

    For each Topic, list the operations the agent needs to perform. Reuse standard Actions where they fit, build custom Actions (wrapping flows, Apex, or Prompt Templates) for the gaps.

  4. Wrap generative work in Prompt Templates

    For Actions that need to generate text, configure the underlying Prompt Template in Prompt Builder. Define grounding, output schema, and constraints.

  5. Pilot, version-control, and deploy through a release pipeline

    Pilot in a sandbox with a small user group. Commit asset metadata to Git. Deploy through change sets or DX, not by editing in production.

Key options
Agent recordremember

The top-level metadata record defining identity, channel, model, and the list of Topics.

Topicremember

The capability-area asset with trigger utterances, instructions, and a list of Actions.

Actionremember

The unit of agent capability: standard (search knowledge, create case) or custom (flows, Apex, Prompt Templates).

Prompt Templateremember

The generative configuration: system instruction, grounding sources, output schema. Reusable across Actions.

Channel bindingremember

The deployment surface: Sales, Service, Slack, messaging. Same Agent can bind to multiple channels.

Gotchas
  • Topics with overlapping trigger utterances confuse the Atlas Reasoning Engine. Keep Topic boundaries sharp.
  • An Action is only as good as its description. The agent decides when to call an Action by reading its description; vague descriptions produce mis-invocations.
  • Agents inherit the assigned user's permissions. An Agent that needs to create a record needs the same access a human user would need; do not skip permission set assignment.
  • Asset packaging behavior is evolving with each Salesforce release. Verify deployment paths against current release notes before committing to a DevOps pipeline.
  • Editing assets in production bypasses version control and breaks the release discipline that keeps agents stable. Treat the Agentforce metadata like flows or Apex.
§

Trust & references

Sources

Cross-checked against the following references.

Official documentation

Straight from the source - Salesforce's reference material on Agentforce Assets.

Was this entry helpful?
Help us write better definitions. Quick reactions or detailed edit suggestions.

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 a key benefit of Agentforce Assets for business users?

Q2. What does Agentforce Assets need to work effectively?

Q3. What technology powers Agentforce Assets in Salesforce?

§

Discussion

Loading…

Loading discussion…