sf-flex-estimator: Agentforce & Data Cloud Flex Credit Estimation

Required Context to Gather First

Ask for or infer:

• agent prompt count by tier: starter, basic, standard, advanced
• action count by type: standard, custom, voice, sandbox
• whether token overages are expected for prompts or actions
• monthly Data Cloud meter volumes, if Data Cloud is in scope
• whether Private Connect is required
• whether the estimate should model a pilot, small production, enterprise, or multiple scenarios
• whether the user wants public list-price guidance or is trying to reconcile contract-specific commercial numbers

If the user does not know exact monthly volumes, start with a baseline template and generate multiple scenarios.

Core Pricing Model

Agentforce

Agentforce billing is linear — no volume tiers.

Data Cloud

Data Cloud uses monthly cumulative tiering.

Other rules

• Flex Credits are priced at $0.004 per FC in this skill.
• Private Connect adds 20% of Data Cloud spend after tiering.
• Agentforce and Data Cloud are estimated separately, then combined.
• Estimates in this skill use publicly documented list pricing only.

For the full meter table and examples, read:

• references/agentforce-pricing.md
• references/data-cloud-pricing.md

Recommended Workflow

1. Baseline the structure

Model the agent and Data Cloud footprint first.

Useful starting templates:

• assets/templates/basic-agent-template.json
• assets/templates/hybrid-agent-template.json
• assets/templates/data-cloud-template.json

2. Calculate the per-invocation cost

For Agentforce, estimate:

per-invocation FC = prompt FC + action FC + token overage FC

3. Calculate Data Cloud base FC

Map each monthly meter volume to the current public rate card, then apply cumulative tiering.

4. Generate scenarios

Use the standard scenario set unless the user provides a better one:

• Low: 1K invocations / month
• Medium: 10K / month
• High: 100K / month
• Enterprise: 500K / month

5. Validate assumptions and recommend optimizations

Check for:

• too many prompts or actions
• unnecessary streaming usage
• likely token overages
• missing Private Connect handling
• unrealistic volume assumptions

Scripts and Templates

Calculator

• assets/calculators/flex_calculator.py
• assets/calculators/tier_multiplier.py

Validation helper

• hooks/scripts/validate_estimate.py

This validator is a manual helper. It is intentionally not wired into the shared auto-validation dispatcher because generic .json or .md file patterns would create too much noise.

Example commands

# Per-invocation estimate for a template
python3 assets/calculators/flex_calculator.py \
  --mode structure \
  --agent-def assets/templates/basic-agent-template.json

# Scenario estimate for an Agentforce + Data Cloud design
python3 assets/calculators/flex_calculator.py \
  --mode scenarios \
  --agent-def assets/templates/hybrid-agent-template.json

# Tiering only
python3 assets/calculators/tier_multiplier.py \
  --base-fc 5000000 \
  --pretty

# Validate an estimate input document
python3 hooks/scripts/validate_estimate.py \
  --input assets/templates/hybrid-agent-template.json \
  --verbose

High-Signal Estimation Rules

• Prefer standard prompts for most production reasoning workloads.
• Use basic prompts only for simple routing/classification.
• Action count often dominates cost faster than prompt count.
• Data Cloud streaming is materially more expensive than prep/query/segment meters.
• Tiering matters only for Data Cloud, not Agentforce.
• Private Connect applies only to Data Cloud spend in this model.
• If the user has contract-specific pricing, treat this skill as a public baseline and note that commercial terms may differ.

Output Format

When the estimate is complete, present:

1. workload summary
2. per-invocation Agentforce cost
3. monthly scenario table
4. Data Cloud tiering impact
5. top optimization recommendations
6. confidence / validation notes

Suggested shape:

Flex Credit estimate: <name>
Agentforce per invocation: <fc> FC ($<cost>)
Data Cloud monthly base: <fc> FC
Scenarios: <low / medium / high / enterprise>
Optimization priorities: <1-3 bullets>
Confidence: <high / medium / low>

Cross-Skill Integration

A common chain is:

sf-ai-agentforce / sf-ai-agentscript / sf-datacloud-* → sf-flex-estimator → sf-deploy

Reference Map

Start here

• README.md
• references/calculation-methodology.md
• references/common-use-cases.md
• references/edge-cases.md

Pricing references

• references/agentforce-pricing.md
• references/data-cloud-pricing.md

Validation and scoring

• references/scoring-rubric.md
• hooks/scripts/validate_estimate.py