Up and running in 10 minutes

FlowMason installs like any other managed package. By the end of this guide you'll have your first AI-powered Salesforce action running — no custom Apex required.

1. Install the package

Run the SFDX command below, or search for FlowMason AI on AppExchange and click Get It Now. Both paths install the same managed package into your target org.

sf package install --package FlowMasonAI --target-org your-org-alias --wait 10

The install takes 2–5 minutes. Once complete, the following are available in your org:

• All FlowMason Apex classes and LWC components
• Custom Metadata Types: FM_Config__mdt, LLMProviderConfig__mdt, FM_Trigger_Binding__mdt, and more
• Custom objects: PipelineExecution__c, Pipeline_Audit__c, Pipeline_Stage_Log__c
• Permission sets: FlowMason_Full_Access, FlowMason_Provider_Admin, FlowMason_Config_Admin
• The Pipeline Studio Lightning app

2. Assign permission sets

Every user who will run or configure FlowMason — including you — needs the right permission set. Go to Setup → Permission Sets and assign the following:

3. Configure an LLM provider

FlowMason ships with Eden AI as the default provider — it's a smart router that picks the best available model for each call. To activate it:

1. Go to Setup → Custom Metadata Types → LLM Provider Config → Manage Records
2. Click the EdenAI Default record (pre-seeded on install)
3. Paste your Eden AI API key into ApiKey__c
4. Ensure IsActive__c is checked
5. Save

If you'd rather use Anthropic, OpenAI, or another provider directly, see the Provider Configuration guide. You can always switch later — that's one of FlowMason's core promises.

4. Verify the install

Open Developer Console → Debug → Open Execute Anonymous Window and paste this. If you see bullet points in the log, you're good.

// Paste this into Developer Console → Execute Anonymous
Account a = [SELECT Id FROM Account LIMIT 1];
System.debug(FMSummarize.of(a.Id).summary);
// Open the log and look for the USER_DEBUG line — bullet points = you're up and running.

5. Run your first AI action

FMSummarize is the first primitive — it generates a concise summary of any Salesforce record. It works on any SObject without per-object configuration: Account, Case, Opportunity, or any custom object you have in your org.

// Option A: one line from any Apex context
String summary = FMSummarize.of(recordId).summary;

// Option B: custom prompt + full response
FMSummarize.Response r = FMSummarize.of(oppId, 'Write a 3-bullet AE briefing');
if (String.isBlank(r.errorMessage)) {
    opportunity.Description = r.summary;
    update opportunity;
}

From Flow Builder, search for AI: Summarize Record and drag it onto the canvas. The action is under the FlowMason AI category. See Invocable Actions for a full walkthrough.

6. Build your first pipeline (optional)

Primitives like FMSummarize cover single-step operations. When you need to chain multiple steps — fetch a record, summarize it, classify the result, then write it back — you use a pipeline. Here's the programmatic API:

String json = new PipelineBuilder('my-first-pipeline')
    .setName('Summarize and notify')
    .addStage('summarize', 'summarize')
        .withConfig('prompt', 'Summarize this record: {{input.recordId}}')
        .end()
    .setOutput('summarize')
    .build();

ExecutionResult result = PipelineRunner.execute(json, new Map<String, Object>{
    'recordId' => someRecord.Id
});
System.debug(result.output);

Or skip the code entirely and build the same pipeline visually in Pipeline Studio. Studio exports SFDX-deployable metadata so you can version-control and promote it across environments.

Troubleshooting

What's next