Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
Sperax × three.ws — a partnership proposal
Historical record: this is a partnership proposal drafted around 2026-07-08, including TVL figures current as of that date. It is a point-in-time pitch, not a status report on shipped work.
TL;DR Sperax has 7 years of shipped DeFi credibility and an auto-yield stablecoin; three.ws has crypto-native attention and a live 3D-agent platform. The trade: they rent our cultural relevance, we rent their credibility and capital. Five pillars: a recurring USDs/SPA-gated AI-credits funnel, USDs-endowed "agents that never die," a 3D guide embedded on Sperax's product via our published SDKs, Sperax's designed agent-economy split going live on our running x402/hiring rails, and a co-marketing sequence where every beat is a real ship, not a press release. 90 days, ordered by what's cheapest to prove first. Ask: a credit-pool budget, a technical contact, and a co-announcement slot.
The trade
Sperax has shipped for seven years, raised $17M from Alameda, Outliers, Polychain, and Amber Group, and built USDs — a stablecoin that pays yield automatically, no staking — but carries low cultural heat for a protocol this established. three.ws has the opposite problem: real trench attention and momentum, on a live agent platform that already does 3D embeds, agent-to-agent payments, and agent hiring in production. Neither side needs to invent trust or attention — we hand each other the thing we already have.
The five pillars
1. Sponsored credits funnel — S
"Hold USDs/SPA, claim monthly AI credits on three.ws" — recurring, not a one-shot airdrop. three.ws already runs a prepaid credit wallet (/credits) that accepts SOL and $THREE deposits and gives $THREE holders up to 30% off every spend; a SperaxOS plugin already exists (/sperax) granting three.ws users free AI credits on chat.sperax.io as a launch offer. This pillar extends that same ledger to accept USDs/SPA and makes the offer recurring. Sperax gets: retention — a reason to keep holding, not just claim once. three.ws gets: qualified, DeFi-native signups every month.
2. Agent Endowments (flagship) — M/L Deposit USDs once; its auto-yield perpetually funds an agent's compute balance on three.ws — "agents that never die." Sperax's own docs describe USDs as earning "by default, no staking required," auto-yield starting "the moment USDs enter your wallet" (usds.sperax.io) — this pillar is new plumbing on a mechanism that already works, wired into three.ws's existing credit ledger, plus a public badge: "self-funding since <date>." Sperax USD carries roughly $521K TVL on Arbitrum (~$551K with staking) per DefiLlama, as of 2026-07-08 — real but early, which is exactly why a flagship new yield sink matters to them. Sperax gets: a genuine new USDs yield sink and the best USDs use-case story in years. three.ws gets: an answer to "who pays for agent compute long-term" that isn't "the user, forever."
3. Their tools, our bodies — S/M
three.ws ships two published, drop-in 3D SDKs: @three-ws/page-agent (a narrating 3D guide + avatar picker, one web component) and @three-ws/tour (a guide that walks a live site, spotlighting and narrating each section — one-tag CDN embed, demoable today at /tour-builder). Either drops onto sperax.io or usds.sperax.io; no new SDK to build. In return, Sperax's DeFi depth — USDs mechanics, veSPA staking, farms — becomes a skill pack three.ws agents call on, matching three.ws's existing 40-pack skill library. Sperax gets: a talking product face, fast. three.ws gets: real DeFi competence for its agents.
4. The agent economy ships on three.ws — L
The rails already run: x402 pay-per-call is live across dozens of endpoints (plus a self-hosted facilitator and the @three-ws/x402-server merchant SDK for turning any endpoint paid), agent-to-agent hiring is live (agent_hire / agent_hire_discover settle real USDC end to end with spend caps and a provenance receipt), and escrowed agent labor already exists via AgenC/Agora — task escrow with competitive and collaborative payout splits, currently in $THREE, plus a royalty-split precedent in the remix-asset settlement rail (creator-set royalty, capped, paid on-chain). Sperax's designed 70/20/10 creator/protocol/staker economics is a policy layer for these rails, not a rail to build from scratch; the concrete new work is adding USDs as a second settlement asset alongside $THREE and USDC. Sperax gets: their agent-economy design actually shipping, on infrastructure that already processes real payments. three.ws gets: a second settlement asset and a credible economic partner.
5. Co-marketing beats — S
Each pillar is a standalone, shippable announcement — not one big launch that slips. Order: credits funnel refresh → 3D guide on sperax.io → endowments pilot → economy split live. Every beat has a working URL to point at.
Sequencing — 90 days
- Weeks 1–2: Pillar 1 (credits funnel extended to USDs/SPA, recurring) and a refreshed
/speraxpage reflecting it. Cheapest to ship — the ledger and the plugin both already exist. - Weeks 3–6: Pillar 3 — a Sperax-tuned tour built with
@three-ws/tourlive onsperax.ioorusds.sperax.io, plus the first Sperax skill pack for three.ws agents. - Weeks 7–12: Pillar 2 pilot (a capped USDs endowment pool, first "self-funding since" badge) and Pillar 4 kickoff — USDs settlement integration and the economy-split design doc, published alongside Pillar 5's next co-marketing beat.
What we need from Sperax
- A credit-pool budget commitment (USDs/SPA) to fund pillar 1's monthly grant and seed pillar 2's endowment pilot.
- A technical contact for the USDs contract interfaces (auto-yield, mint/redeem) and a plugin review on the SperaxOS side.
- A co-announcement slot for the week 1–2 beat.
Both sides already ship for a living — this proposal only asks each of us to do more of what we're already good at, pointed at each other.