Skip to content

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.

docs(growth): draft Coinbase Founders Fuel and Solana Foundation grant applications with verified evidence

nirholas4854eb47ff8f
diff --git a/marketing/growth/applications/gated/coinbase-founders-fuel.md b/marketing/growth/applications/gated/coinbase-founders-fuel.md
new file mode 100644
index 000000000..47e55fa05
--- /dev/null
+++ b/marketing/growth/applications/gated/coinbase-founders-fuel.md
@@ -0,0 +1,180 @@
+# Coinbase Developer Platform Founders Fuel: application package
+
+**Status: draft, not submitted. Commit-gated.** This file names crypto projects other than $THREE (Coinbase, Base, USDC, x402 facilitators), so under the CLAUDE.md commit gate it stays uncommitted until the owner approves the content. Nothing here has been sent to Coinbase.
+
+- Opportunity row: `marketing/growth/opportunities.csv`, "Coinbase CDP Founders Fuel", target date 2026-09-23.
+- Campaign row: `marketing/growth/campaigns.csv`, `MKT-2026-10-CDP-X402`, anchor date 2026-10-13.
+- Program page: https://www.coinbase.com/developer-platform/discover/launches/founders-fuel
+- Application form: https://forms.gle/YDpPXdPQ5sYR5QGy7 ("Founders Fuel Interest Form")
+- Evidence gathered: 2026-09-17, against production revision `three-ws-api-00438-cg2` (commit `477eb6f8c`, from `https://three.ws/api/version`).
+
+---
+
+## 1. What the program asks for (verified 2026-09-17)
+
+From the program page:
+
+- **Who it is for.** Builders who "Have a functioning app or working product", "Plan to launch publicly within 3 months", "Are integrating one or more CDP tools (Onramp, CDP Wallet(s), x402, etc.)", and "Show a clear roadmap and intent to ship". Bonus points for onboarded users or early traction.
+- **What selected teams receive.** Up to $15,000 in CDP and Paymaster credits; meetings with CDP product leads; a Launch Spotlight ("Social amplification + inclusion in the CDP Builder Showcase"); GTM support (positioning, onboarding UX, partner introductions if aligned); beta access to new developer tools; $5,000 in AWS Activate credits.
+- **How to activate.** Apply through the form, share a live product link, submit a launch plan ("just 1 to 2 paragraphs"), and say which credits and tools you want. Review takes 30 to 60 days. A Telegram group is linked for questions.
+
+Consequence for timing: a 2026-09-23 submission may not be decided before the 2026-10-13 campaign anchor. The launch plan below does not depend on acceptance; acceptance amplifies it.
+
+---
+
+## 2. What is actually live (the evidence, with how it was checked)
+
+Present Solana first, truthfully: three.ws is a Solana-native platform, and x402 on Solana runs through its own self-hosted facilitator. The Coinbase Developer Platform and Base lane is an additional, already-wired rail. The application asks Coinbase to help us take that additional rail public, not to replace the home rail.
+
+### 2.1 Solana: the home rail (production proof)
+
+| Claim | Evidence | How verified |
+|---|---|---|
+| A self-hosted x402 facilitator is live and settles Solana USDC with no third party in the settlement path | `GET https://three.ws/api/x402-facilitator` returns `"service":"three.ws self-hosted x402 facilitator"`, `x402Version: 2`, `scheme: exact`, network `solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp`, with `/supported`, `/verify`, `/settle`, `/discovery/resources` | curl, 200 |
+| Volume since 2026-07-03 | 136,534 successful settles and 932,590 successful verifications recorded in `x402_self_facilitator_log`; 2,832.536 USDC moved by successful settles | read-only DB query |
+| Last 30 days | 8,594 successful settles from 3 distinct payer wallets | read-only DB query |
+| Signed receipts | 67,263 rows in `x402_receipts`; buyers retrieve theirs at https://three.ws/receipts ("Receipt Vault") | DB count, page 200 |
+| Third-party attribution | x402scan lists the facilitator at https://www.x402scan.com/facilitator/three-ws since `Merit-Systems/x402scan` PR #1032 ("Add three.ws self-hosted Solana facilitator") merged 2026-08-11 | curl 200, `gh pr view` |
+| Code | `api/_lib/x402/self-facilitator.js`, `api/x402-facilitator/[action].js`, Apache-2.0 at https://github.com/nirholas/three.ws | repo |
+
+**Honesty note that must stay in any submitted text.** The Solana settlement volume is overwhelmingly the platform's own autonomous canary loop (`docs/x402-ring-economy.md`, `docs/autonomous-x402.md`): platform-controlled payer wallets buying platform endpoints to prove every paid route settles end to end. In the last 30 days no successful settle came from a wallet outside the ring wallet table. It proves the rails work at volume; it is not external customer demand. Do not describe it as customer traction.
+
+### 2.2 Coinbase Developer Platform and Base: the additional rail
+
+| Claim | Evidence | How verified |
+|---|---|---|
+| Official CDP SDKs are production dependencies | `@coinbase/x402` `^2.1.0` and `@coinbase/cdp-sdk` `^1.50.0` in `package.json`; `createCdpAuthHeaders` from `@coinbase/x402` authenticates facilitator calls in `api/_lib/x402-spec.js`; `generateJwt` from `@coinbase/cdp-sdk/auth` signs Onramp session requests in `api/_lib/coinbase-onramp.js` | code |
+| Base settles route through the CDP facilitator whenever CDP credentials are present | `facilitatorFor()` in `api/_lib/x402-spec.js` returns the CDP facilitator for Base and the other CDP EVM networks when `CDP_API_KEY_ID` and `CDP_API_KEY_SECRET` are set | code |
+| Those credentials are set in production | `CDP_API_KEY_ID` (literal) and `CDP_API_KEY_SECRET` (Secret Manager reference) are on the `three-ws-api` Cloud Run service | `scripts/read-service-env.mjs --names` |
+| Every priced endpoint offers a Base USDC accept beside the Solana ones | `https://three.ws/.well-known/x402.json` lists 4,521 resources; every one carries a Solana USDC accept, and Base (`eip155:8453`) accepts appear 9,041 times across them | curl + parse |
+| The 402 challenge carries the Bazaar discovery extension | Live `GET https://three.ws/api/x402/crypto-intel` answers `402` with accepts ordered Solana USDC, Solana $THREE, then Base USDC, and `extensions` = `bazaar`, `eip2612GasSponsoring`, `erc20ApprovalGasSponsoring`, `payment-identifier`, `offer-receipt`; the base64 `PAYMENT-REQUIRED` header mirrors it | curl |
+| Real Base mainnet settlements have landed | Two settles on 2026-09-05 (`/api/x402/skill-marketplace`, tx `0xe86b82b06b96e320e60ee1bbbc8fa322b950b41e7241f417f93f5a61955bf4c7`; `/api/x402/solana-register-health`, tx `0xfb98ae4b90ef71929bda91719613546a2a434b3fc0eda9910b8e3835f85d1068`), 0.001 USDC each; the first receipt reads `status 0x1` in Base block 50,888,803 | `x402_audit_log` + Base RPC `eth_getTransactionReceipt` |
+| A Bazaar browser consumes the CDP discovery API | https://three.ws/bazaar ("Search and browse the x402 facilitator catalog. Filter by network, price, and extensions. Pay any service in one click.") backed by `api/_lib/x402/bazaar-client.js`, which reads `https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources` alongside other facilitators; `/api/bazaar/list` currently returns 998 HTTP resources and `/api/bazaar/providers` 1,146 providers | curl, 200 |
+
+### 2.3 The gap we will not hide
+
+On 2026-09-17 a full scan of the CDP Bazaar discovery API (15,733 items, paginated 1,000 at a time) returned **zero** resources on `three.ws`. The Bazaar indexes an endpoint only after its first verify plus settle passes through the CDP facilitator, and drops it after 30 days without a settle (`docs/x402-distribution.md`). Only two real Base settles have landed, and neither is currently listed.
+
+This is exactly the "working product, not yet launched on this rail" position Founders Fuel is built for, and the application should say so plainly rather than claim a Bazaar listing that does not exist.
+
+Closing the gap is already scripted: `scripts/x402-bazaar-settle-sweep.mjs` pays one real Base USDC settle per live endpoint, cheapest first, under a spend cap (default $3), using each endpoint's own runnable `bazaar.input` example. It needs a funded Base buyer wallet (`X402_BUYER_PRIVATE_KEY`), which exists in neither `.env` nor `.env.local` nor the Cloud Run service today. Running it is a spend action and needs the owner's explicit yes (CLAUDE.md gate 1).
+
+---
+
+## 3. Form answers (copy-ready)
+
+Fields marked **OWNER** need facts only the owner has. Everything else is drafted from verified evidence.
+
+| Field | Draft answer |
+|---|---|
+| Full Name | **OWNER** |
+| Email Address | **OWNER** (partnership mail routes to [email protected] per `docs/partners.md`; the form likely wants a person) |
+| Telegram | **OWNER** |
+| LinkedIn | **OWNER** |
+| Project or Startup Name | three.ws |
+| Location (City, Country) | **OWNER** |
+| Team Size | **OWNER** |
+| Website or Demo Link | https://three.ws/bazaar |
+| GitHub / Code Repository | https://github.com/nirholas/three.ws |
+| Which CDP products | x402 (primary) and Onramp. Onramp is real code: `api/onramp/link.js` mints CDP Onramp session tokens through `generateJwt` from `@coinbase/cdp-sdk/auth` (`api/_lib/coinbase-onramp.js`), enabled by the same CDP credentials, and the live endpoint answers `GET https://three.ws/api/onramp/link` with a Solana-address validation error, so it defaults to funding Solana wallets. Do not tick CDP Wallet API or Embedded Wallets: no code for them was found. |
+| Describe your product in one sentence | three.ws is a production platform where AI agents create, inspect, and pay for 3D work over x402, with every paid call priced in one discovery catalog and every settlement backed by a signed receipt. |
+| What problem are you solving? Who is your target audience? | Agent builders need tools their agents can find and pay for without an account, an API key, or a human in the loop. three.ws turns 3D generation, rigging, model inspection, and agent data into priced HTTP endpoints an agent discovers, pays, and retries in one round trip. The audience is developers building autonomous agents and MCP clients who want creative and 3D capabilities they can buy per call. |
+| Is this project live? | Yes, live product |
+| How are you planning to use Coinbase Developer tools? | Our x402 rail runs Solana first through our own facilitator; the CDP facilitator is our Base lane. Every priced endpoint already advertises a Base USDC accept, emits the Bazaar discovery extension in its 402 challenge, and routes Base verify and settle through the CDP facilitator with `@coinbase/x402` auth. The next step is getting every endpoint indexed in the Bazaar by settling through CDP, keeping those listings fresh with recurring settles, and publishing a reproducible "agent buys 3D work" builder demo on that lane. |
+| What stage is your company or project at? | **OWNER** (choose from Solo builder / Pre-seed / Seed / Post-seed) |
+| What would Founders Fuel help you accomplish in the next 30 to 60 days? | Take the Base lane from wired to indexed: every three.ws endpoint listed in the CDP Bazaar with clean metadata, a public builder demo where an agent discovers a 3D capability, receives an HTTP 402, pays, and gets its result plus a receipt, and CDP product-lead feedback on our Bazaar metadata before the 2026-10-13 public session. |
+| What kind of support are you most interested in? | Technical support; Co-marketing; Feature prioritization / roadmap feedback. Product credits only if the owner confirms CDP usage will be billed. |
+| Open to being featured in Coinbase developer content? | Yes (**OWNER** to confirm) |
+| Final notes | Short version of section 2, including the honest Solana-first and ring-volume framing, plus the live links: https://three.ws/bazaar, https://three.ws/receipts, https://three.ws/.well-known/x402.json, https://three.ws/x402/studio. |
+
+### Launch plan (the "1 to 2 paragraphs" the program page asks for)
+
+> three.ws already runs a live x402 economy: 4,521 priced endpoints in a public discovery catalog, a self-hosted Solana facilitator that has recorded more than 136,000 successful settlements since July (largely our own autonomous canary loop proving every route end to end), and signed receipts for every call. Every endpoint also advertises a Base USDC accept routed through the Coinbase Developer Platform facilitator, and two Base mainnet settles have landed. What we have not done is launch that lane publicly: our endpoints are not yet in the CDP Bazaar.
+>
+> Over the next 90 days we will settle every endpoint through CDP so the Bazaar indexes it, keep those listings fresh with recurring real settles, and publish a reproducible builder demo on 2026-10-13 in which an agent discovers a 3D capability, receives an HTTP 402, pays in USDC on Base, and receives its result and receipt, with an integration guide other agent builders can copy. We will measure settled CDP calls, indexed endpoints, and distinct external buyers, and report them publicly at day 30, 60, and 90. Founders Fuel fits because the remaining work is distribution and feedback, not prototype risk.
+
+Before pasting: re-run the counts in section 5 and replace every number with that day's value (templates.md rule: never paste numbers from an old article).
+
+---
+
+## 4. 90-day launch plan (consistent with MKT-2026-10-CDP-X402)
+
+Campaign row fields this plan honors: anchor proof `https://three.ws/bazaar`; primary channel X; secondary channels Coinbase builder channels, GitHub Discussions, Telegram community, three.ws news; audience CTA "Run one paid call and inspect its receipt"; partner ask "Founders Fuel Launch Spotlight and Builder Showcase review"; primary KPI "settled CDP calls and partner response"; status `awaiting_application`.
+
+Day 0 is the submission date, 2026-09-23. Spend actions are marked **(gate 1)** and need the owner's explicit yes each time; posting to external channels is marked **(gate 2)**.
+
+### Days 0 to 20 (2026-09-23 to 2026-10-13): from wired to indexed
+
+1. Submit the form (owner) and join the program Telegram group for questions.
+2. Fund a dedicated Base buyer wallet and set `X402_BUYER_PRIVATE_KEY` locally **(gate 1, owner decides amount)**.
+3. `node scripts/x402-bazaar-settle-sweep.mjs --dry-run`, review the endpoint list and spend, then run it under `--max-usd=3` **(gate 1)**. All Base accepts pay the owner's wallet, so the spend is an internal transfer plus facilitator-paid gas.
+4. Verify indexing with the CDP discovery API scan in section 5 (not the `EXTENSION-RESPONSES` header, which upstream issue x402-foundation/x402#2112 documents as sometimes missing).
+5. Fix any endpoint the Bazaar rejects (most likely cause: a `bazaar.input` example that does not validate against its `inputSchema`, or zero-amount auth-hint accepts; see "Known listing blockers" in `docs/x402-distribution.md`).
+6. Record the builder demo: an agent calls `/api/x402/crypto-intel` or a 3D endpoint, receives the 402, pays on Base through CDP, gets the 200 plus `X-PAYMENT-RESPONSE`, and the receipt is visible at `/receipts`. Record the Solana path in the same session so the home rail leads the piece.
+
+### Day 20 (2026-10-13): public anchor
+
+7. Publish the campaign piece "An AI agent discovers and purchases 3D work over HTTP 402" on X, with the Solana path shown first and the CDP Base lane shown as the second rail **(gate 2)**.
+8. Cross-post to GitHub Discussions, the Telegram community, and three.ws news **(gate 2)**; add a `data/changelog.json` entry if the Base lane indexing is a user-visible change.
+9. Share the demo and integration guide in Coinbase builder channels; if accepted by then, request the Launch Spotlight and Builder Showcase review **(gate 2)**.
+
+### Days 20 to 60 (to 2026-11-22): keep listings alive and earn ranking
+
+10. Keep every indexed endpoint inside the Bazaar's 30-day window with at least one real CDP settle per endpoint per 30 days. The Solana ring payer (`api/_lib/x402/pay.js`) is Solana-only, so this is a recurring run of the sweep script **(gate 1, owner sets the monthly cap)** until a Base leg is added to the autonomous loop.
+11. Publish the integration guide as a docs page under `docs/` (the Base lane section of `docs/x402-buyer.md` or a new tutorial), with a runnable example.
+12. Day 30 public metrics post **(gate 2)**. If accepted, use the CDP product-lead session for Bazaar metadata and ranking feedback.
+
+### Days 60 to 90 (to 2026-12-22): conversion and report
+
+13. Target outreach to agent and MCP builders who already buy on the Bazaar, with the one-paid-call CTA.
+14. Day 60 and day 90 metrics posts **(gate 2)**, reporting the same KPIs, with external buyers separated from internal sweep and ring payers.
+15. Update `marketing/growth/campaigns.csv` status and the opportunity tracker with the application ID, decision, and any Builder Showcase or social URL.
+
+### KPIs and how each is measured
+
+| KPI | Source | Day 0 baseline (2026-09-17) |
+|---|---|---|
+| three.ws resources indexed in the CDP Bazaar | Full scan of `https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources` | 0 of 15,733 |
+| Settled CDP calls (Base) | `x402_audit_log` rows `event_type='payment_settled'`, `network='eip155:8453'`, excluding the two `/api/x402/replay-test` fixture rows | 2 |
+| Distinct external Base buyers | Same rows, payer not the sweep buyer wallet | 0 |
+| Solana settles, 30 days (home rail, reported first) | `x402_self_facilitator_log` `action='settle' and ok` | 8,594 from 3 payers |
+| Partner response | Form decision, Launch Spotlight or Builder Showcase URL | none |
+
+---
+
+## 5. Re-verification commands (read-only, run the day you submit)
+
+```bash
+# Production revision
+curl -s https://three.ws/api/version
+
+# Facilitator and catalog
+curl -s https://three.ws/api/x402-facilitator
+curl -s https://three.ws/.well-known/x402.json | node -e "let s='';process.stdin.on('data',d=>s+=d).on('end',()=>{const r=JSON.parse(s).resources;const base=r.filter(x=>x.accepts.some(a=>a.network==='eip155:8453')).length;console.log({resources:r.length,withBaseAccept:base})})"
+
+# Live 402 challenge: accepts order and extensions
+curl -s https://three.ws/api/x402/crypto-intel | node -e "let s='';process.stdin.on('data',d=>s+=d).on('end',()=>{const j=JSON.parse(s);console.log(j.accepts.map(a=>a.network),Object.keys(j.extensions||{}))})"
+
+# CDP Bazaar presence (full paginated scan)
+for off in $(seq 0 1000 16000); do curl -s "https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources?limit=1000&offset=$off"; echo; done | grep -o '"resource":"https://three.ws[^"]*"' | sort -u | wc -l
+```
+
+Database counts come from `x402_self_facilitator_log`, `x402_audit_log`, and `x402_receipts` with `DATABASE_URL` from `.env.local` (SELECT only).
+
+---
+
+## 6. Wording rules for anything sent
+
+- Solana leads every paragraph that mentions chains. Base is "an additional rail", never "our chain".
+- Never call ring or sweep settles "customers" or "traction". Say "autonomous canary loop" or "internal settles".
+- Never claim a Bazaar listing until the section 5 scan returns a non-zero count that day.
+- Partner statuses, exactly as `docs/partners.md` states them: "OpenAI Select Partner"; "IBM Business Partner" (with the rule that the public `/api/ibm/*` tools are not IBM partnership deliverables); "member of NVIDIA Inception"; "member of Google Cloud for Web3 Startups"; "AWS Partner, and the Marketplace listing is coming"; "member of the Quicknode Startup Program". No endorsement is claimed by any of them.
+- Do not pitch $THREE price. $THREE may appear only as the second Solana accept that is factually in the challenge.
+
+## 7. Owner-only inputs before submission
+
+1. Personal fields: name, email, Telegram, LinkedIn, location, team size, company stage.
+2. Approval to commit this file (commit gate: names projects other than $THREE).
+3. Whether to fund a Base buyer wallet, how much, and the monthly settle cap for keeping listings alive (gate 1).
+4. Confirmation that Onramp is a lane the owner wants to present (it is wired and credentialed, but no end-to-end purchase was run for this draft).
+5. Approval of every external post in section 4 (gate 2), and consent to be featured in Coinbase developer content.
diff --git a/marketing/growth/applications/gated/solana-foundation-grant.md b/marketing/growth/applications/gated/solana-foundation-grant.md
new file mode 100644
index 000000000..f2b3878e5
--- /dev/null
+++ b/marketing/growth/applications/gated/solana-foundation-grant.md
@@ -0,0 +1,224 @@
+# Solana Foundation grant proposal: sponsor-safe, self-hostable x402 facilitation for Solana
+
+**Status: draft, not submitted. Commit-gated.** This file names crypto projects other than $THREE (the Solana Foundation, USDC, other x402 facilitators and packages), so under the CLAUDE.md commit gate it stays uncommitted until the owner approves the content. Nothing here has been sent to the Solana Foundation.
+
+- Opportunity row: `marketing/growth/opportunities.csv`, "Solana Foundation public-good grant", status `scope_needed`, target date 2026-10-16.
+- Program page: https://solana.org/grants-funding
+- Application form: https://share.hsforms.com/1GE1hYdApQGaDiCgaiWMXHA5lohw ("Solana Foundation Funding Application")
+- Form guidance: https://docs.google.com/document/d/1eK-WNhQmQFyk06XhwWRZJrXn3z_0NDbChZ_Rm8XRoXw
+- Evidence gathered: 2026-09-17.
+
+---
+
+## 1. What the Foundation asks for (verified 2026-09-17)
+
+- **Funding types.** Milestone-based grants for public goods, convertible grants for public goods with a commercial component, and Requests for Proposals. This proposal is a **milestone-based public-good grant**.
+- **Public good definition.** A project that "either makes a significant open-source contribution to the Solana ecosystem, or if it has a meaningful free community offering."
+- **Required.** A project overview, how it delivers public value, "a well structured budget proposal, and include thoughtful milestones", open-sourced learnings, and an answer to "Why Solana?"
+- **Milestones.** The guide asks for "measurable milestones for both Product Development and Adoption", each unlocking a stated portion of funding.
+- **Process.** Rolling review (about one week), optional due diligence (about three weeks to a decision), then legal finalization. No typical amount is published.
+- **Form fields.** Company name, website, country, legal first and last name, email, Solana on-chain accounts, funding amount, funding category, the project, open-source status, amount plus milestones, usage metrics, competition, public good, why you.
+
+---
+
+## 2. Choosing the scope: one package, one gap
+
+### Candidates considered
+
+All are Apache-2.0 and live in this repository. Download counts are npm's own API (`api.npmjs.org/downloads/point/last-month`, window 2026-08-13 to 2026-09-11, fetched 2026-09-17). npm counts include mirrors and CI installs, so they indicate reach, not users.
+
+| Package | What it is | npm last month | Why not chosen (or chosen) |
+|---|---|---|---|
+| `@three-ws/metaplex-agent-mcp` 0.2.1 | MCP server that registers agents in the Metaplex Agent Registry | 944 | Highest reach, but a wrapper over an existing registry; the grant would fund convenience, not a missing primitive |
+| `@three-ws/vanity` 0.1.3 | Solana vanity address grinder | 374 | Well-served category already |
+| `@three-ws/skill-license` 0.1.2 | Skill licenses as SPL NFTs plus a PDA | 366 | Narrow audience today |
+| `@three-ws/onchain-agent-wallets` 0.1.1 | Agent spending allowance enforced by the SPL Token program | 356 | Strong idea, little production evidence yet |
+| `@three-ws/solana-agent` 0.2.2 | Solana agent SDK (keypair, wallet adapter, transfers, x402 exact) | 333 | Broad SDK; hard to scope one verifiable public good |
+| **`@three-ws/x402-server` 0.1.3** | **Merchant side of x402: issue the 402, verify, settle, fee split; Solana leads** | **228 (1,559 from first publish 2026-06-24 to 2026-09-16)** | **Chosen: backed by the platform's largest body of production evidence, and the grant closes a verified gap (below)** |
+
+### The gap (verified against code)
+
+1. **Every Solana x402 merchant depends on someone else's facilitator.** `@three-ws/x402-server` settles through a facilitator URL, and its `DEFAULT_FACILITATOR_URL` is a hosted third-party service (`packages/x402-server/src/index.js`). The same is true of most x402 merchant tooling.
+2. **three.ws already runs its own Solana facilitator in production, but it is not reusable.** `api/_lib/x402/self-facilitator.js` (951 lines) plus `api/x402-facilitator/[action].js` verify and settle Solana USDC with the platform's own fee payer over its own RPC. It reads platform env vars, the platform connection helper, and a platform payTo allowlist, so nobody else can adopt it.
+3. **The reference Solana facilitator does not protect the sponsor's SOL balance.** The upstream `@x402/svm` exact facilitator (x402-foundation/x402, `typescript/packages/mechanisms/svm/src/exact/facilitator/scheme.ts`, read 2026-09-17) validates the transaction shape well: it rejects a fee payer that transfers funds, caps compute-unit price, checks mint, amount, recipient, and memo. It contains no sponsor balance floor, no fee budget over time, and no minimum settle amount. Those are the failures that actually stopped settlement in production here (section 4).
+
+### The scope, in one sentence
+
+**Add a self-hostable, sponsor-safe Solana facilitator to `@three-ws/x402-server`, extracted from the production three.ws facilitator and built on the upstream `@x402/svm` transaction checks, so any Solana merchant can verify and settle x402 payments with their own fee payer and cannot be drained or griefed; contribute the sponsor guards and the attack tests upstream.**
+
+Deliberately out of scope: EVM chains, new token programs, hosted services, dashboards beyond one public metrics endpoint, and anything involving $THREE.
+
+---
+
+## 3. Why Solana
+
+- **Fee-payer sponsorship is native.** A Solana transaction names its fee payer, so the facilitator co-signs as fee payer and the buyer needs no SOL, only USDC. That is what lets an AI agent with nothing but a stablecoin balance pay per call. The same property is the risk this grant fixes: the fee payer signs the whole transaction, so a careless facilitator can be drained (the production code's own security note documents this vector).
+- **Per-call pricing survives the fee.** Production sets a minimum sponsor-mode settle of 1,000 atomic USDC ($0.001) because each co-sign burns about 5,000 lamports (`MIN_SPONSOR_SETTLE_ATOMIC` in `self-facilitator.js`). A tenth-of-a-cent API call that still covers its own settlement cost is the economics x402 needs.
+- **SPL `TransferChecked` plus Memo** gives a strict, decodable instruction set the facilitator can validate exactly before signing.
+
+---
+
+## 4. Production evidence (how each number was checked)
+
+| Claim | Evidence | Method |
+|---|---|---|
+| The facilitator is live | `GET https://three.ws/api/x402-facilitator` returns `"service":"three.ws self-hosted x402 facilitator"`, `x402Version: 2`, `exact` on `solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp`, fee payer `WwwuGbqHrwF5RG89KhUbmRWEvjnRH9k5kVM5p7T3WwW` | curl, 200 |
+| Lifetime volume (since 2026-07-03) | 136,534 successful settles, 932,590 successful verifications, 2,832.536 USDC moved | read-only query of `x402_self_facilitator_log` |
+| Last 30 days | 8,594 successful settles from 3 payer wallets | same table |
+| Guards fire in production | 775,408 refused settles; the largest reasons are `fee_runway_exhausted` (fee budget cap) and `fee_wallet_below_floor` (sponsor floor), plus `payer_fee_unfunded` and simulation failures on verify | same table, grouped by `reject_reason` |
+| Independent listing | x402scan lists it at https://www.x402scan.com/facilitator/three-ws since `Merit-Systems/x402scan` PR #1032 merged 2026-08-11 | curl 200, `gh pr view` |
+| Tested | 71 facilitator test cases across `tests/x402-self-facilitator-{settleable,floor-accounting,fee-meter,min-settle,settle-recovery}.test.js`, `tests/api/x402-facilitator-dispatch.test.js`, `tests/api/x402-facilitator-log-durable.test.js`, `tests/api/healthz-self-facilitator.test.js`; the package's own suite passes 25 of 25 (`node --test`, run 2026-09-17) | test files, local run |
+| Package published | `@three-ws/x402-server` versions 0.1.0 (2026-06-24) through 0.1.3 (2026-09-11), zero runtime dependencies | npm registry |
+
+**Honesty note that must stay in any submitted text.** The settlement volume is the platform's own autonomous canary loop (`docs/x402-ring-economy.md`): platform-controlled payer wallets paying platform endpoints, with a single platform payTo, to prove every paid route settles end to end. In the last 30 days no successful settle came from a payer outside the ring wallet table. The numbers prove the facilitator and its guards under sustained load. They are not third-party adoption, and the proposal must not describe them that way. The large refusal count is the guards doing their job under a deliberately tight fee budget, and should be explained, not hidden.
+
+---
+
+## 5. Milestones
+
+T0 is the date the grant agreement is signed. Every deliverable is public and checkable by the Foundation without access to three.ws systems. Each milestone releases the share of funding shown in section 7.
+
+### M1. Standalone facilitator module (development), T0 + 3 weeks
+
+- Move the verify and settle path out of platform code into `packages/x402-server/src/facilitator/`, published as the subpath export `@three-ws/x402-server/facilitator`. The zero-dependency core stays zero-dependency; Solana libraries load only through the subpath.
+- Every platform coupling becomes an explicit option: RPC connection or URL list (failover), fee-payer signer, payTo allowlist, log sink.
+- Transaction-shape validation delegates to the upstream `@x402/svm` exact facilitator checks wherever they cover the case, instead of duplicating them (open source first); three.ws-specific checks stay only where upstream has no equivalent.
+- The production three.ws endpoint switches to the package, so the package is what runs in production from this milestone on.
+
+**Verifiable deliverable:** npm release with the subpath export; `README.md` with a runnable example that stands up a facilitator in one file; the existing 71 facilitator tests ported to the package and passing in `node --test`; `https://three.ws/api/x402-facilitator` still answering, now served by the published version (version string exposed in the response).
+
+### M2. Sponsor guards (development), T0 + 6 weeks
+
+Ship as documented, individually configurable guards, all proven in production today:
+
+- **Sponsor SOL floor:** refuse to settle when the fee payer is below a configured balance (production default 0.02 SOL, `SPONSOR_SOL_FLOOR_LAMPORTS`).
+- **Fee budget meter:** cap lamports burned per rolling window (`fee_runway_exhausted` in production).
+- **Minimum sponsored settle:** refuse dust transfers that cost the sponsor more in fees than they move (`MIN_SPONSOR_SETTLE_ATOMIC`).
+- **Recipient allowlist:** co-sign only transfers to merchant-controlled accounts.
+- **Rent-failure detection and bounded ATA-create rent**, so a new recipient token account cannot become an open-ended rent drain.
+- **Idempotent settle and settle recovery:** a retried settle never double-broadcasts, and a broadcast whose confirmation was lost is reconciled rather than repeated.
+- **Durable decision log interface:** every verify and settle, with a machine-readable refusal reason.
+
+**Verifiable deliverable:** npm release; one test per guard; a `docs/` page in the package explaining each guard, its default, and the failure it prevents.
+
+### M3. Adversarial test suite and upstream contribution (development plus ecosystem), T0 + 9 weeks
+
+- An attack suite of signed-but-malicious Solana transactions run against the facilitator: fee-payer lamport drains, priority-fee inflation, dust griefing, recipient substitution, mint substitution, duplicate submission, lost-confirmation replay.
+- The same suite packaged so any x402 Solana facilitator can run it against its own `/verify` and `/settle` endpoints.
+- A pull request to `x402-foundation/x402` adding opt-in sponsor-balance and fee-budget hooks (or equivalent the maintainers prefer) plus the attack cases to `@x402/svm`, with an issue opened first to agree the design.
+
+**Verifiable deliverable:** the suite published in the package with a CLI entry that targets any facilitator URL; the upstream issue and PR links, public. Funding for M3 releases on the PR being opened with passing upstream CI and a maintainer review requested. A merge is not required, because the Foundation should not pay on a third party's schedule; the merge outcome is reported in M5.
+
+### M4. Reference deployment and operator guide (adoption), T0 + 12 weeks
+
+- A one-command self-host path: a `Dockerfile` and an example service in the package, deployable to any container host.
+- An operator guide: generating and funding a fee payer, choosing floors and budgets, RPC failover, monitoring, and what each refusal reason means.
+- A public read-only metrics endpoint on the three.ws deployment reporting settles, verifications, refusals by reason, and sponsor lamports burned, split into internal canary traffic and external payers.
+- Registration of the reference deployment in public x402 facilitator directories where one exists.
+
+**Verifiable deliverable:** the Dockerfile and guide in the repository; the metrics endpoint live; a recorded walkthrough of a fresh facilitator settling a mainnet USDC payment in under 15 minutes from `npm install`.
+
+### M5. Adoption report (adoption), T0 + 6 months
+
+**Verifiable deliverable:** a public report with the section 8 metrics against the section 8 baselines, the upstream PR outcome, every issue filed by outside operators and how it was resolved, and security fixes shipped during the maintenance window.
+
+---
+
+## 6. Timeline
+
+| Week | Work |
+|---|---|
+| T0 to T0+3 | M1 extraction, production cut-over to the package |
+| T0+3 to T0+6 | M2 guards and per-guard docs |
+| T0+6 to T0+9 | M3 attack suite, upstream issue and PR |
+| T0+9 to T0+12 | M4 self-host path, operator guide, metrics endpoint |
+| T0+12 to T0+26 | Maintenance, operator support, upstream follow-through, M5 report |
+
+---
+
+## 7. Budget (proposal, every figure owner-adjustable)
+
+**This table is a starting proposal for the owner to adjust, not a quote.** The Foundation publishes no typical grant size, and the guide asks that the amount be "intentional and explained in the milestones section". Engineering is costed at an assumed blended rate of **$5,000 per engineer-week**; the owner should replace that rate with the real one before submitting. The security review figure is an estimate to be replaced with a real vendor quote.
+
+| Line item | Milestone | Basis | Amount (USD) | Rationale |
+|---|---|---|---|---|
+| Facilitator extraction and production cut-over | M1 | 2 engineer-weeks | 10,000 | Decouple env, RPC, signer, allowlist, and log sink; port 71 tests; switch production to the package |
+| Sponsor guards and per-guard docs | M2 | 3 engineer-weeks | 15,000 | Seven guards, each generalized from platform code, tested, and documented |
+| Attack suite and upstream contribution | M3 | 2 engineer-weeks | 10,000 | Malicious-transaction corpus, facilitator-agnostic CLI, upstream design discussion and PR iterations |
+| Self-host path, operator guide, metrics endpoint | M4 | 2 engineer-weeks | 10,000 | Container path, guide, public metrics split internal versus external |
+| Independent security review of the co-signing path | M2 to M3 | external review, estimate | 10,000 | The fee payer signs whole transactions; a second set of eyes on the validation and guards is the difference between "tested" and "safe to recommend" |
+| Mainnet conformance funds | M3 to M4 | SOL for sponsor fees and ATA rent, USDC for test payments | 500 | Attack and conformance runs on mainnet, not only devnet; unspent funds reported in M5 |
+| Maintenance and operator support | M5 | 2 engineer-weeks spread over 14 weeks | 10,000 | Issue triage, security fixes, upstream review follow-through |
+| **Total** | | | **65,500** | |
+
+Proposed release schedule (also adjustable): M1 15%, M2 25%, M3 20%, M4 20%, M5 20%.
+
+---
+
+## 8. Ecosystem impact and success criteria
+
+**Who benefits.** Any Solana developer who wants to charge per request over HTTP 402 without routing settlement through a third party; any existing facilitator operator who wants the sponsor protections and the attack suite; `@x402/svm` users, if the upstream contribution lands.
+
+**Baselines (2026-09-17), then targets.** Targets are goals the grant is measured against, not claims of current adoption.
+
+| Metric | How measured | Baseline | Target by M5 |
+|---|---|---|---|
+| Independent operators running the facilitator on mainnet | A public `/supported` endpoint or directory listing that identifies the package, confirmed by the operator | 0 | 3 |
+| External (non-canary) payers settling through the three.ws reference deployment, 30 days | M4 metrics endpoint | 0 | 25 |
+| `@three-ws/x402-server` npm downloads, last month | `api.npmjs.org/downloads/point/last-month` | 228 | 1,000 |
+| Public repositories depending on the package | GitHub dependency search for `@three-ws/x402-server` outside `nirholas/*` | owner to record at submission | 5 |
+| Attack cases in the suite, all passing on the package | Suite output in CI log | 0 | at least 20 |
+| Upstream contribution | `x402-foundation/x402` PR state | none | opened with review requested (M3); outcome reported (M5) |
+| Security review findings | Review report | not reviewed | report published, all high and critical findings fixed |
+
+**Learnings shared openly.** The operator guide, the attack corpus, the security review summary, and the M5 report are all public in the repository under Apache-2.0.
+
+---
+
+## 9. Form answers (copy-ready)
+
+| Field | Draft answer |
+|---|---|
+| Company name | three.ws |
+| Website URL | https://github.com/nirholas/three.ws/tree/main/packages/x402-server |
+| Country | **OWNER** |
+| First Name / Last Name | **OWNER** (legal name; the guide says no pseudonyms) |
+| Email Address | **OWNER** |
+| Solana On-Chain Accounts | WwwuGbqHrwF5RG89KhUbmRWEvjnRH9k5kVM5p7T3WwW (facilitator fee payer), wwwwwDxFWRn7grgr3Esrsg5C6NvDoDHSA4gaCffccrU (facilitator payTo). Both verified from `GET https://three.ws/api/x402-facilitator` on 2026-09-17. |
+| Funding Amount | **OWNER** (proposal: $65,500, see section 7) |
+| Funding category | **OWNER** to pick from the live dropdown; the closest fit is payments or developer tooling. Not DeFi. |
+| Your project / idea | Section 2 "The scope" sentence, then section 2 "The gap" in three short paragraphs. Not referred by a Foundation partner unless the owner says otherwise. |
+| Open sourced? | Yes (Apache-2.0) |
+| Amount and milestones | Section 5 milestones with the section 7 release percentages expressed in dollars |
+| Usage metrics | Section 4 table, including the honesty note, plus the npm numbers in section 2 |
+| Competition | Hosted x402 facilitators that support Solana (for example the public PayAI facilitator, whose `/supported` lists Solana mainnet on 2026-09-17) are convenient but put a third party in the settlement path. The upstream `@x402/svm` reference facilitator is self-hostable and validates transaction shape, but has no sponsor balance floor, fee budget, or minimum-settle guard. This package is self-hosted like the reference and adds the sponsor-economics guards and an attack suite from a facilitator that has run on mainnet since July 2026, and it contributes those back upstream rather than competing with the reference. |
+| Public good | Apache-2.0 code, zero-dependency core, a facilitator-agnostic attack suite any operator can run, an operator guide, and an upstream contribution to the shared reference implementation. No token, no fee, no account required to use any of it. |
+| Why you | Three months of operating a mainnet Solana facilitator under continuous load: 136,534 settles and 932,590 verifications logged, the guards in section 5 written in response to real failures in that log, 71 facilitator tests, an independent listing on x402scan. **OWNER** to add team background and any Solana hackathon history. |
+
+Do not add website links to answers whose question does not ask for one (form instruction).
+
+---
+
+## 10. Re-verification commands (read-only, run the day you submit)
+
+```bash
+curl -s https://three.ws/api/x402-facilitator
+curl -s https://three.ws/api/x402-facilitator/supported
+for p in x402-server metaplex-agent-mcp vanity skill-license onchain-agent-wallets solana-agent; do
+ curl -s "https://api.npmjs.org/downloads/point/last-month/@three-ws/$p"; echo
+done
+(cd packages/x402-server && node --test test/*.test.js)
+gh pr view 1032 -R Merit-Systems/x402scan --json state,mergedAt
+```
+
+Settlement counts come from `x402_self_facilitator_log` with `DATABASE_URL` from `.env.local` (SELECT only), filtering `ok` by `action`, and excluding payers in `x402_ring_wallets` for the external-payer figure.
+
+## 11. Owner-only inputs before submission
+
+1. Legal first and last name, email, country (the form requires a real name).
+2. The real engineering rate and final funding amount, replacing the section 7 assumption; a real security review quote.
+3. Funding category selection from the live dropdown.
+4. Team background and any Solana hackathon history for "Why you".
+5. Whether any Foundation partner or staff referral exists.
+6. Approval to commit this file (commit gate: names projects other than $THREE).
+7. Agreement that the honesty note on canary-loop volume appears in the submitted text.