Pricing your agents
A practical guide to setting prices that earn real money at scale — without pricing yourself out of the market.
How pricing works
When you publish agents on Envelope, you set the price deployers pay every time your agents run. Envelope retains a fixed 10% platform fee and pays the remaining 90% to you via Stripe Connect.
Per run — Charged each time the agents are invoked, regardless of token count. Best for agents that complete a defined, bounded task.
Per 1K tokens — Charged based on how many tokens the run consumes. Best for open-ended tasks where output length varies significantly.
Recommended ranges
Per-run pricing
| Complexity | Agents | Price range |
|---|---|---|
| Lightweight | 1–2 agents, short tasks | $0.01 – $0.05 / run |
| Mid-complexity | 3–5 agents, chained reasoning, tool use | $0.05 – $0.25 / run |
| Heavy | Long-context, many agents, multiple rounds | $0.25 – $1.00+ / run |
Per-1K-token pricing
Platform compatibility — Per-token billing requires the deployment platform to report exact token counts on each run. Currently supported on Paperclip. Relevance AI does not expose token counts via its API — teams deployed there will not generate billing events under this model.
| Markup | Price range | Notes |
|---|---|---|
| 2–5× model cost | $0.002 – $0.01 / 1K tokens | Where most API resellers land |
The 3–5× rule of thumb
Charge 3–5× your underlying model API cost per run.
If a 5-agent run costs you ~$0.01 in Claude or GPT API calls, pricing at $0.04–$0.05/run gives you healthy margin while staying cheap enough that deployers don't think twice about it.
Managed runtime agents: price on value, not compute — If you use the HTTP adapter in managed mode, Envelope runs a hosted model on your behalf — and covers that inference cost as part of its 10% platform fee. Your effective underlying cost is near zero. Price toward the value the agents deliver, not your compute spend.
The Envelope sweet spot
For most agents, $0.05–$0.10 per run is the sweet spot:
- Enough to earn real money at scale — 400 runs/month = $18–$36 after Envelope's fee
- Low enough that a Paperclip company would rather pay it than rebuild the agents themselves
- Competitive with the cost of calling the underlying models directly
Thinking about deployer value
Price based on what the agents save the deployer, not just what it costs you to run.
A sales outreach agent that books one extra meeting per month is worth far more than $0.05. A customer support agent that deflects 10 tickets is worth far more than $0.10. If your agents deliver clear, repeatable value, price toward the higher end of the range.
Revenue at scale
| Price | 100 runs/mo | 400 runs/mo | 1,000 runs/mo |
|---|---|---|---|
| $0.05 / run | $4.50 | $18.00 | $45.00 |
| $0.10 / run | $9.00 | $36.00 | $90.00 |
| $0.25 / run | $22.50 | $90.00 | $225.00 |
Figures shown after Envelope's 10% platform fee.
Setting pricing via API
Include pricingModel and priceUsdCents in the definition when creating or patching a team:
PATCH /templates/:id{
"pricingModel": "per_run",
"priceUsdCents": 10
}pricingModel | Description |
|---|---|
"free" | No charge to deployers |
"per_run" | Fixed charge per run trigger |
"per_k_tokens" | Charge per 1,000 tokens consumed across the run |
priceUsdCents is the amount in US cents charged to deployers. Envelope's 10% platform fee is deducted automatically before payout. Only draft status templates can be patched — create a version draft first if the team is already published.
Price changes applied to a new version take effect for new runs immediately after publishing. Existing installs continue running under their previously agreed price until the deployer upgrades.
Receiving payouts
Envelope pays builders via Stripe Express. Connect your account once from your Account page — Stripe handles identity verification, tax forms, and bank deposits automatically. Payouts are batched per billing cycle and typically settle within 2–7 business days.