GuidesAgentForce guides
AgentForce guides
Task-by-task walkthroughs for using ready-made agents and building, operating, and scaling your own custom agents.
These guides show you how to get work done with agents — from chatting with a ready-made catalog agent to building and operating your own. New to agents? Start with Use ready agents. Ready to build your own? Jump to Build your first agent.
Using agents is open to every member. Building and editing agents is admin-only.
For how each feature works under the hood, see the AgentForce reference.
Learning path
Use ready agents
- Meet the catalog — pick the right agent
- Have your first agent conversation
- Get better results from an agent
Build your first agent
- Plan a custom agent
- Create your agent's identity & model
- Write effective instructions
- Give your agent tools
- Test before you publish
- Publish your first agent
Operate & scale
- Schedule an agent to run on its own
- Version & roll back an agent
- Read analytics & improve an agent
- Design a multi-agent setup
- Assistant vs. catalog vs. custom — choose the right worker
Browse by topic
Get started with agents
Meet the catalog
Pick the right ready-made agent for the job.
Your first agent conversation
Use suggested prompts and get a useful first result.
Threads & files
Manage conversations and attach files.
Get better results
Prompt an agent so it nails the task.
Build your first agent
Plan a custom agent
Decide when to build and scope the role.
Identity & model
Name it, style it, pick its model.
Write effective instructions
Define how the agent behaves.
Test before you publish
Validate a draft in the test panel.
Publish your first agent
Ship a version to your whole org.
Tools & integrations
Give your agent tools
Let it take actions, not just talk.
Connect to your apps
Wire in Gmail, Slack, GitHub, and more.
Add MCP & external tools
Extend an agent with MCP servers.
Scope what it can do
Set guardrails on agent actions.
Knowledge & memory
Give your agent knowledge
Ground it in your KB and skills.
Configure agent memory
Control what it remembers across chats.
Automate & schedule
Schedule an agent
Run it on a recurring schedule.
Trigger from events
Kick it off when something happens.
Test, version & improve
Version & roll back
Publish versions and restore earlier ones.
Analytics & improve
Use usage data to make it better.
Troubleshoot an agent
Diagnose a misbehaving agent.
Manage & scale
Start from a template
Build faster from a starting point.
Manage agents for your org
Enable, disable, and track the 20-agent cap.
Personality & brand voice
Make agents sound like your brand.
Design a multi-agent setup
Split work across specialized agents.
Choose the right worker
Assistant vs. catalog vs. custom.