Methodology
This page documents every formula AgentMargin uses, so you can verify results independently.
Shared formulas
- Total cost = fixed costs + variable costs + labor + other allocated costs.
- Gross profit = revenue − direct costs.
- Gross margin % = (gross profit ÷ revenue) × 100.
- Markup % = (profit ÷ cost) × 100.
- Price for target gross margin = total cost ÷ (1 − target margin).
- Break-even clients = monthly fixed overhead ÷ contribution per client, rounded up.
- Contribution per client = (retainer + usage revenue) − variable cost per client − payment fees.
Cost models by billing unit
Providers bill differently, so each input is priced in its own unit and never assumed to include anything else: per 1M tokens (input/output priced separately), per minute (voice, telephony, transcription), per 1,000 characters (TTS), per number per month, and flat monthly fees. Model price never includes telephony, voice, hosting or other services.
Voice agent model
Calls × duration gives minutes; the inbound share splits telephony between inbound and outbound rates. The AI model is priced either per token (tokens per call × calls ÷ 1,000,000 × price per 1M, input and output separately) or per minute. Voice generation is priced per 1,000 characters or per minute. Agent-platform fees are per minute and per-call fees per call. Phone numbers and hosting scale with agents (clients × agents per client); storage and tools are per client.
Custom AI agent model
Each cost line has its own unit: per minute (× total minutes), per interaction (× total calls or conversations), per 1M tokens or per 1,000 characters (× quantity per interaction × interactions), per client, per agent, or flat monthly.
Cost per call
Cost per call = call minutes × per-minute stack (model + voice + transcription + telephony + platform) + per-call fees + fixed monthly costs ÷ calls per month.
Chatbot model
Conversations × messages gives message count; messages × tokens per message gives token totals, priced separately for input and output. Support cost = hours × hourly rate. Embeddings, vector DB, storage, hosting and external APIs are per-client monthly figures.
Agency profit model
MRR = clients × (retainer + usage revenue). Setup fees are one-time revenue and are excluded from gross profit, which uses recurring revenue only. Operating profit additionally subtracts overhead, payment fees and one-time implementation cost. Gross profit and estimated operating profit are always labelled distinctly.
Pricing model
Risk buffer = percentage of base cost, applied before margin: price = (base cost + buffer) ÷ (1 − target margin) in margin mode, or × (1 + markup) in markup mode. Target margin is capped below 100% so the formula can never divide by zero.
Simulation model
Scenarios multiply expected usage (defaults 0.5× low, 1× expected, 2× high, plus a custom multiplier — all editable). Variable cost scales with minutes; fixed costs and revenue stay constant — which is exactly why high usage compresses margin.
Data integrity
- No fake or random results: outputs come only from your inputs or labelled samples.
- Published rates carry a source URL and last-verified date; unverified rates, or rates older than 90 days, are marked "Needs verification".
- Negative inputs are clamped to zero; division-by-zero cases return warnings, not misleading numbers.
- Core formulas are covered by automated unit tests with hand-verified expected values.
Provider pricing sources
Sample provider rates live in a separate configuration file that an administrator updates without touching calculator code. Every rate stores provider, service, billing unit, price, currency, source URL, last-verified date and effective date. No model price is assumed to include telephony, voice, hosting or any other service. Every price can be overridden in the calculators.
| Provider | Service | Sample rate | Status |
|---|---|---|---|
| OpenAI | GPT-4o — input tokens source | $2.5 per 1M input tokens | Needs verification |
| OpenAI | GPT-4o — output tokens source | $10 per 1M output tokens | Needs verification |
| OpenAI | GPT-4o mini — input tokens source | $0.15 per 1M input tokens | Needs verification |
| OpenAI | GPT-4o mini — output tokens source | $0.6 per 1M output tokens | Needs verification |
| Anthropic | Claude Sonnet — input tokens source | $3 per 1M input tokens | Needs verification |
| Anthropic | Claude Sonnet — output tokens source | $15 per 1M output tokens | Needs verification |
| Anthropic | Claude Haiku — input tokens source | $0.8 per 1M input tokens | Needs verification |
| Anthropic | Claude Haiku — output tokens source | $4 per 1M output tokens | Needs verification |
| Google AI | Gemini 1.5 Flash — input tokens source | $0.075 per 1M input tokens | Needs verification |
| Google AI | Gemini 1.5 Flash — output tokens source | $0.3 per 1M output tokens | Needs verification |
| ElevenLabs | Text-to-speech (conversational tier) source | $0.15 per 1,000 characters | Needs verification |
| Deepgram | Speech-to-text (Nova family) source | $0.0043 per minute | Needs verification |
| OpenAI | Whisper API transcription source | $0.006 per minute | Needs verification |
| Twilio | Local number — inbound voice source | $0.0085 per minute | Needs verification |
| Twilio | Local number — outbound voice (US) source | $0.014 per minute | Needs verification |
| Twilio | Local phone number rental source | $1.15 per number / month | Needs verification |
| Telnyx | Inbound voice (local number) source | $0.005 per minute | Needs verification |
| Telnyx | Outbound voice (US local number) source | $0.0075 per minute | Needs verification |
| Telnyx | Local phone number rental source | $1 per number / month | Needs verification |
| Vapi | Platform fee per minute (additional to providers) source | $0.05 per minute | Needs verification |
| Retell AI | Platform fee per minute (additional to providers) source | $0.07 per minute | Needs verification |
See also the Disclaimer and Privacy Policy.
AgentMargin provides estimates based on the information and pricing assumptions entered by the user. Provider pricing, taxes, usage, and service terms may change. Verify current charges with each provider before setting client prices. This tool does not guarantee profitability.
AgentMargin provides estimates based on the information and pricing assumptions entered by the user. Provider pricing, taxes, usage, and service terms may change. Verify current charges with each provider before setting client prices. This tool does not guarantee profitability.