The three-tier model

Attribution first. Plutus measures; it does not save money. Perseus (routing / resolve-before-context) and Perseus Vault (memory) are what reduce spend. Plutus meters the spend and proves the savings. So the savings-share is an ecosystem monetization — it only applies when Perseus is in the loop producing (and Plutus proving) real savings. Standalone, Plutus is a "track my own spend, verify I'm getting my tokens" meter, and there's nothing to share.

Plutus has one primitive: a verifiable efficiency meter that records what your AI stack would have cost on a flagship model versus what it actually cost, on a tamper-evident hash chain. Everything below is a way to monetize that one number for a different segment. The key design idea: savings-share is a single lever, set to a different position per tier.

Tier	Who	Price	Reporting	Savings-share
Free	Individual devs	$0, unlimited metering	Headline savings number only	Suggested — optional tip jar
Pro	Power users / prosumers	$20/mo flat	Full depth + chain-verifiable ledger	Waived — the flat fee replaces it
Team	Companies	$10/seat/mo	Full depth + team attribution	Mandatory — 10% of provable savings
Enterprise	Large orgs	Custom	Everything + SSO/SLA/self-host	Negotiated

Why three tiers and not one price

They map to willingness-to-pay and to how provable the value is:

1. Free is the growth engine, not the revenue. Metering is unlimited so the efficiency billboard keeps running for real workloads. Standalone it reads "Your AI spend: $X · flagship-equivalent$Y · N× efficient" (a tracking / verification stat); with Perseus in the loop it becomes "Perseus saved you $X — verified by Plutus" — every free user a walking advertisement for their own number. The tip jar ("chip in a share of what Perseus saved you") earns little directly; its job is virality + goodwill + a funnel into Pro/Team. The number is honest because it's tamper-evident and reconstructable — that's the moat.

2. Don't double-dip on individuals. Charging a solo user $20/mo and a % of their savings feels predatory and kills conversion. On Pro the flat $20 is the deal — predictable, no variable bill. Savings-share is waived.

3. Savings-share is a business instrument. Companies expect outcome-based pricing and their savings numbers are large enough to justify a %. Team pays a per-seat floor (covers attribution + admin) plus a mandatory 10% of the provable savings. This is the only tier where the share is billed by default.

Where the lever lives in the code

• plutus_agent/pricing.py — Tier.savings_share ∈ {suggested, waived, mandatory, custom, none}, Tier.per_seat_usd_month, Tier.full_reporting, and pricing.savings_mode(tier_key).
• Dashboard billboard — server/app._dashboard injects the current-month efficiency + savings_share into the summary; server/views.render_dashboard branches on whether a Perseus baseline is present (`savings_share.covered_events

  0): ecosystem → "Perseus saved you $X — verified by Plutus"; standalone → "Your AI spend … flagship-equivalent … N× efficient" (no "saved" claim). The Free tip jar (POST /billing/checkout/donate`) shows only when there are provable savings to share.

• Reporting gate — Tier.full_reporting hides the per-task / leakage / export panels on Free behind a "Pro feature" card.
• Billing guard — plutus bill-savings --apply refuses to invoice a pro (waived) or free (suggested) org unless --force; mandatory billing is the Team path.

Donations (the Free tip jar)

donate_checkout opens a one-time Stripe Checkout (kind=donation). On checkout.session.completed the webhook records a distinct donation ledger entry — hash-chained like everything else, never confused with prepaid credit or a savings-share invoice. On the hosted instance the payment is real revenue; the matching account credit is a thank-you.

Not yet automated (follow-ups)

• Per-seat Team subscriptions — Team checkout is currently "talk to us"; the per-seat Stripe subscription + seat roster/roles are a follow-up. Savings-share for Team is billed today via plutus bill-savings --apply (mandatory tier, no --force needed).
• Aggregate + individual attribution UI for Team.