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.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
beat: progress log
Append one dated entry per finished or partially finished order: order number, commit SHAs, what remains and who owns it.
- 2026-10-09: campaign created (27 orders, 601 to 627). Nothing run yet.
- 2026-10-23: order 603 (pay CLI: keychain, x402, MPP) done. Source modules landed in 2b7bb3921 (swept by another agent's commit); tests, commands, README, docs and changelog in the commit titled feat(cli): three-ws pay. Remaining, owner-gated: the first real mainnet settlement (gate 1), command in the final report.
- 2026-10-23: order 612 (prepaid balances settled on-chain in batches) done. Ledger, batcher, route hook (
paidEndpoint({ mode: 'prepaid' })), endpoints, crons and migration landed in a37074e67 (swept by another agent's commit); the interval fix, constants split, reconcile script, triage entry and docs landed in ed75c5946 (also swept by another agent); this commit closes the order and logs it. Remaining, owner-gated: apply migration 20261023160000 (npm run db:statusfirst), fund the settler and setX402_PREPAID_SETTLE_LIVE=1for the first mainnet batch settlement (gate 1), details in the final report. - 2026-10-23: order 605 (owner-funded on-chain allowance budgets) done except the real devnet run: faucet 429 (owner or faucet is the dependency); the same lifecycle script passed on a local validator running the real program. Code landed in ed75c5946 (swept by another agent's commit); docs, pages, changelog and STRUCTURE row in the commit titled docs(wallet): document owner-funded on-chain agent budgets. Run
node scripts/allowance-devnet-proof.mjswhen devnet SOL is available. - 2026-10-23: order 602 (outcome-weighted memory and chained decision trail) implemented and proven against the live dev database: clean chain verifies ok, a manual edit fails at exactly the edited index, a loss-tied memory ranks below an equal win-tied one and is still returned. Code, migration 20261023150000 (applied alone via --file after db:status), tests, docs, STRUCTURE row and changelog landed in 22a5a1f49 and 2b7bb3921 (swept by other agents' commits). The per-agent hash chain, verify route and badge already existed, so no backfill was needed. Order file left in place: the full
npm testandnpm run audit:docscould not be run cleanly here (broken local node_modules install; audit polluted by node_modules.bak). Owner: reinstall deps, then run both and delete the order file if green. - 2026-10-23: order 604 (agent self-onboarding) done except the first real mainnet upgrade payment (gate 1, owner). Challenge-response signup, migration, limits and service landed in 2b7bb3921 (swept by another agent's commit); the upgrade route, CLI
signup --agentand onboarding tests in ed75c5946 (swept); the absolute resource URL fix, route and upgrade tests, docs, skill update, pages, STRUCTURE and changelog in the commit titled docs(onboarding). Proven on a local server with a fresh keypair: signup, one account per key, a printed 402 for Builder (29.00 USDC) and a simulated settlement that grants the plan once. - 2026-10-23: order 601 (skills from experience) done. Detector, distiller, refinement, dashboard list, API, MCP tool, migration and tests landed in 22a5a1f49, 2b7bb3921 and d93b25d5b (swept by other agents' commits); docs, API reference, STRUCTURE row, changelog and the template-written failure section in 265074a13. Proven against the live dev database with a real model: run -> candidate row -> valid draft -> listed -> accepted -> next run called the installed tool -> dismissed -> nothing installed. Nothing remains for the owner except pushing.
- 2026-10-23: order 605 (on-chain allowance budgets) PARTIAL. Program client, service and tests are in place; the devnet proof run (scripts/allowance-devnet-proof.mjs) is blocked by a devnet faucet 429. Order file left. Owner: rerun the proof once the faucet recovers or fund the proof keypair, then delete the order.
- 2026-10-23: order 606 (burner wallet with owner confirmation) PARTIAL. Tier classifier, server guard, approval stamp, owner wallet signature path, decision-trail writes (kind burner_gate), the Burner cap tab, docs, STRUCTURE row and changelog landed (27d81c33f and earlier swept commits); 26 burner tests pass. Not verified: full
npm test,audit:docs, and a devnet-agent run. Order file left. Owner: run those three and delete the order if green. - 2026-10-23: order 617 (transaction landing race) done. Provider table, race, per-provider stats, failure-streak skip with probes and per-trade tip ceiling landed in d9cac5c7f, 18e7ef17f, 60a700988, 6666b1cad, b988762ac (swept by other agents' commits); flush serialization, the tiny-transaction compute-budget fix, devnet script and the data/pages.json entry in the closing commit. The proof run used a local solana-test-validator because the public devnet faucet returned 429; migration 20261023190000 is pending apply.
- 2026-10-23: order 620 (backtest integrity and research/live wall) done in bfd8ca754 (the canary cron entry in vercel.json landed earlier in ab9e36b55). Look-ahead audit and prior-causality check run on every sniper backtest, dataset manifests with content hash (old results marked unreproducible), 30 minute price and candle canaries ordering the SOL price and chart chains, ResearchContext plus a transitive import-wall test. Verified: 19 tests in tests/backtest-integrity.test.js, wall demonstrated failing on a planted execution-engine import then reverted, live canary run against 11 real sources, audit:docs clean. Owner: sync the new cron to Cloud Scheduler at the next deploy; tests/cron-scheduler-sync.test.js fails on a stale documented cron count (docs/build.md says 145, vercel.json has 150) that other orders also move.
- 2026-10-23: orders 602, 611 and 613 closed. 602 (outcome-weighted memory and chained trail) landed in 22a5a1f49 and 2b7bb3921; 611 (credit plans and metered MCP tools: plan tables, dispatcher meter with debit, refund, trials and -32402 fallback, /mcp-plans, docs/mcp.md pricing) in a37074e67 and 22a5a1f49; 613 (agent inbox: code_senders migration, extractor, inbox_list, inbox_read, inbox_wait_for_code, docs/agent-mail.md) in 22a5a1f49. Migrations show applied in db:status. Touched-area tests pass (mcp-credit-plans 12, agent-inbox-otp and agent-inbox-e2e, memory and ledger suites); the full vitest run still has about 50 failing files from other lanes in flight (news_resolve, presetScopes, pay_and_call, grok docs), none in these areas.
- 2026-10-23: order 618 (chat trade commands with typed confirmation) done in 41264073598. /buy, /sell, /snipe, /orders, /cancel behind an exact-yes 60 second confirmation, tp/sl exits behind GATEWAY_TPSL_ORDERS, PNG P&L cards, trade section in docs/chat-gateways.md, STRUCTURE row, changelog. Real mainnet trade not run (gate 1).
- 2026-10-23: order 607 (x402 v2 upto scheme) done in 744292da3 and 3c4ab8722: upto on the LLM proxy, 402 body shows the scheme, handler error settles nothing, exact clients unchanged. tests/x402-upto.test.js passes.
- 2026-10-23: order 608 (MPP on Solana) done in c28db0aa. One URL answers both x402 and MPP Payment challenges, buyerFetch pays either with caps and a dryRun quote, 15 tests in tests/mpp-solana.test.js including CLI interop and a cap block, docs/mpp-solana.md. Owner (gate 1): real mainnet settlement is not run; command is
three-ws pay https://three.ws/api/x402/gas-oracle --max 0.002. - 2026-10-23: order 609 (one definition, many rails) done in 8d56cd92. Dry-run quote is identical on x402, mpp, a2a and ap2 (10000 / 9750 / 250), disabling a rail 404s, 22 tests in tests/skill-rails.test.js, docs/skill-rails.md. The publish UI rail checkboxes already existed.
- 2026-10-24: order 624 (installable skills repository) done. Spec validator (
npm run check:skills-spec), clean-directory install test (npm run test:skills-install,npx skills addlists and installs 20/20, 6/6, 14/14), 21 skills' frontmatter fixed,npx skills addinstall line in the generated repo READMEs, listing copy in docs/skill-directory-listings.md. Remaining, owner-gated (gate 2): publishing the three repos and posting listings; commands print fromnpm run build:skill-repos. Commit: see git log subject 'feat(skills)'. - 2026-10-24: order 622 (Solana agent registry) built, PARTIAL. Registration beside registerIdentityV1, registry block on the agent row, live tier read and minRegistryTier guard in a2a-hire, tier tie-break in offer ranking, agent-page badge, rate-limited back-fill (scripts/register-8004-agents.mjs), docs/solana-agent-registry.md, 15 unit tests. Round trip (register, feedback, tier read) proven on a local validator that cloned the devnet registry programs. Remaining: rerun scripts/prove-8004-registry.mjs against public devnet once the faucet limit (429) clears (owner: any session with a funded devnet key); mainnet back-fill is gate 1 (SOLANA_8004_REGISTER_MAINNET=1 and --confirm). Commit: see git log subject 'feat(solana)'.
- 2026-10-24: order 623 (skill supply-chain scanner) done. acorn AST scanner over handlers.js, tools.json, SKILL.md and manifest (requires.network host declarations), verdicts clean/review/blocked enforced on publish, edit and candidate publish, admin moderation queue, publish-form Check skill with file, line and fix guidance, 39 tests over a planted-malicious corpus. Catalog of 152 skills rescanned: 152 clean, 0 review, 0 blocked. docs/skill-supply-scan.md.
- 2026-10-24: order 626 (metered OpenAI-compatible router) built, PARTIAL, commit fbb235df5. Unkeyed POST /v1/chat/completions answers 402 with an x402 upto challenge sized to the model worst case, settles measured tokens times price (streaming close and dropped streams included, upstream failure cancels the deposit), prices served from /v1/models, 9 tests in tests/metered-router.test.js, docs/metered-router.md. Remaining: npm run audit:docs is not clean because of other agents' open items (owners: those orders); a real paid request is gate 1 (owner, funded wallet). Order file left in place.
- 2026-10-24: order 610 (hire escrow evaluator and disputes) PARTIAL, implementation complete. Machine, ledger, evaluation, sweep cron (vercel.json), migration, docs/labor-market.md and changelog were already shipped; 4 labor suites pass (102 tests: lifecycle pass, fail, dispute, timeout, property test, audit chain). Full vitest run: 47 files fail elsewhere (unrelated in-flight work, e.g. x-content-approval, mcp-studio-ratelimit), none in labor, burner or approvals. Order file left. Owner: get the full suite and
audit:docsgreen, then delete orders 606 and 610.