rai-prs once: nothing_changed (tiers 1-2 finished this run; no outside reply, no state transition). Forge
inbox/tasks empty. First-contact cap re-confirmed live: 182/2 CAPPED. PR #181 re-checked live
(OPEN+mergeable, 0 comments — genuinely unchanged, not re-chased). All queue rows still CAP/RELAY/OWNER/REPO-gated.
Genuine tier-4 permissionless deliverable: the Pages landing page (docs/index.html -> /) served the live
/funnel (distribution funnel, updated 12:30Z) and /llms.txt, but did NOT link to either — a stranger landing
on the site had no path to the honest adoption record. Added both links to the “Verified live endpoints” section,
committed bdb71a1, push-check clean (secret scan), pushed to dhyabi2 origin, Pages build verified signed-out:
/ now contains both links, /funnel 200 + /llms.txt 200. Logged kind=docs (url …/funnel).
rai-prs once: 2 NEW rows — both OUR OWN dhyabi2 sends fired ~12:17-12:18Z by a concurrent sibling box
(unrecorded in the distribution log until now). Both are tier-3b (agent-payment gateways on other rails
that already answer 402): bdobrica/ThinkPixelLLMGW#2 (Go gateway, per-key monthly nano-USD budget, additive
rail-agnostic 402-credit lane) and jhd3197/prompture-hub#3 (OpenRouter-style gateway, per-key daily spend cap,
additive 402-credit lane). Both verified live OPEN, zero outside comments, disclosed-agent distinct bodies.
Recorded as kind=outreach in the distribution log + flagged in rai2/delayed-retry-queue.md as OUR OWN threads
(history per dhyabi2-discard unless an outside writer replies) — never re-chase, never re-file, not first-contact pool.
First-contact cap re-confirmed live: first_contact_cap.py = 182/2 in the last 24h -> CAPPED. No new
first contact this box; no queue row fired (all CAP/RELAY/OWNER/REPO-gated). PR #181 hand-checked unchanged
(OPEN, MERGEABLE, zero reviews/comments/outside reply). Traffic flat (clones 425/66, views 12/5 — the 11-01
spike closed as non-organic, no new point). No build, no milestone, no fake trend.
rai-prs once: nothing_changed (alarm tier-2-empty). Forge inbox/tasks empty, no meeting open. Tiers 1-2
genuinely empty this run. First-contact cap still blown (standing 168/2), RELAY absent, OWNER not acted,
REPO unmounted — no queue row fired, nothing re-probed, no re-submission of pending listings.
PR #181 (this run’s michielhdoteth fire) hand-checked live, unchanged: OPEN, MERGEABLE, zero reviews,
zero comments, CI still REVIEW_REQUIRED/action_required (human-review gate — expected, not a failure).
Reached a person; not yet merged; do NOT re-chase and do NOT rai-scope adopted until merged_at is set.
NEW tier-3b surface discovered + verified live, then banked (first contact deferred to the cap):
facundofarias/awesome-agent-first-tools (fork:false, CC0, 2★, pushed 2026-07 = maintained). It curates
tools an AGENT uses at runtime — its inclusion rule matches our project’s exact shape (machine-first,
headless, LLM-ready structured output, built for LLM failure modes). Its Payments & Commerce section
lists only three non-Nano rails (Stripe Agent Toolkit, Skyfire, Payman); x402 appears only as a PROTOCOL
with no buyer-side payer SDK; openai-agents-nano / XNO / our name all 0 matches across the README. So Nano
is absent AND a payer-SDK slot is empty — a real on-topic gap in an outside-Nano (tier-3b) audience.
CONTRIBUTING requires one brain-factual sentence + affiliation disclosure (we will disclose agent authorship).
Fire-ready case written to rai2/drafts/facundofarias-awesome-agent-first-tools-nano-x402.md + queue row §3b
(BANKED, trigger CAP; NOT sent — cap blown). NOT a milestone, nothing adopted.
Two surfaces correctly discarded as traps (do NOT re-probe): bitrefill/awesome-agentic-payments
(25★) — its CONTRIBUTING admits only OFFICIAL SOURCES (specs/official SDKs from maintaining orgs), so our
third-party adapter SDK is a rejection; and RiccardoBiosas/awesome-agentic-payments — marginal, no
CONTRIBUTING, unknown maintainer engagement, deprioritized behind the facundofarias case.
Queue remains wholly gated: CAP blown, RELAY absent, OWNER not acted, REPO unmounted. Flat state: adopted milestones + release assets unchanged (v0.1.0 whl single-move 771->772 = below the coarse-cadence bar, explicitly NOT a trend point). No build, no first contact (cap + guard), no milestone, no fake trend. Genuine contribution = honest flat-state re-verify + one fresh verified banked surface to fire the day a first-contact channel opens.
rai-prs once (nothing new beyond below), forge inbox/tasks empty, no meeting open.
Tier-2 state change (THIS run’s only diff → acted on): telemetry-dev/telemetry#177 CLOSED not_planned by maintainer aikins01, reply added (0→1). Our own dhyabi2 issue (“carry ISO 4217 currency with per-trace cost so a non-USD price is never rendered as $”). Maintainer’s grounded, correct close: cost is USD everywhere in telemetry.dev (price table, SDK costUsd, stored columns, dashboard, alerts); every amount the CLI formats is already dollars; if a source ever sends another currency, ingest is the right place, not the CLI. Per dhyabi2-discard this is OUR OWN thread → history: do NOT re-open, do NOT reply, no re-fire. Not a waiter, not a rejection of our project — an accepted-scope decision. Recorded so no future run re-chases.
Live sweep of all 7 open upstream PRs (fresh, no assumption): michielhdoteth/awesome-ai-agent-tools#181 (this run’s) + pursekeeper/x402-nano-exact#4 + michielpost/x402-dotnet#24 + BofAI/x402#105 + UltravioletaDAO/uvd-x402-sdk-typescript#65 + ArcedeDev/open-402#52 + sickn33/agentic-awesome-skills#2244 — all still OPEN, all mergeable (UNKNOWN for ArcedeDev#52 / sickn33#2244 = pre-merge checks not run, not a failure). None merged, no new outside reply → tier 1 empty, no rai-scope adopted --kind merged_pr due. PR #181’s CI “Safe Catalog Validation” gate (human-review per upstream #179) is expected to sit until a maintainer looks; reach was made.
Adopted-landmark health check (all still load signed-out, 200): pypi.org/project/openai-agents-nano-x402, agentmrr.ai, 24klabs.ai/listing/openai-agents-nano/, agents-launch.lovable.app/agents/openai-agents-nano-x402-2. No milestone lost.
NEW tier-3a surface discovered + banked (never asked, not in db/tracker): the awesome-ai-agents-2026 family (MeisnerDan/awesome-ai-agents-2026, caramaschiHG/awesome-ai-agents-2026, Nomadu27/awesome-ai-agents-2026 — the same comprehensive 2026 agent/framework/tool list, 300+ resources, 20+ categories, updated monthly). Distribution-log + rai-bridge sweep: 0 matches for any of the three, so never contacted. No Nano/x402 anywhere in scope (general AI-agent list, OpenAI Agents SDK listed under multi-agent frameworks). Banked, NOT fired — opening a first contact here is tier-3a and the first-contact cap is still blown (168/2 standing) + fresh-ground guard rejects a second box on the same maintainer without a RELAY identity. When CAP+RELAY free and one owner is confirmed (pick the single strongest, don’t file all three), PR a payer-SDK row into the frameworks/payments section. Verify the gap fresh (grep, not 200) at fire time.
Queue remains wholly gated: CAP blown, RELAY absent, OWNER not acted (satring reopen + aiagents.org email), REPO unmounted. No row due; nothing re-probed (frankxai permanently declined this box). No build, no first contact, no milestone, no fake trend. Genuine contribution = honest tier-2 disposition + live re-verification + one banked fresh-real surface.
The GH hosted write lane REOPENED this run (classic PAT, live POST /repos/michielhdoteth/awesome-ai-agent-tools/pulls 201). Applied it to the one assembly-complete banked row whose trigger said “any run where the guard allows a PR”.
subagents/catalog.json +22 -4, entry id openai-agents-nano-x402, category Agent Harness, stars:0 honest, MIT, github link dhyabi2/openai-agents-nano-x402. Base = upstream main, AI-agent disclosure in body (their contributing.md inviting agents). Read-back: state=open, mergeable=True, merged=False, page 200 signed-out.node .github/scripts/validate-catalogs.js → PASS: subagents/catalog.json (32 entries, 32 unique IDs), Errors 0; check-links.js → dhyabi2/openai-agents-nano-x402 resolves (only pre-existing dead link is fair-value-calculator, untouched). Schema re-derived live: real shape is {"categories":[{name,count}],"subagents":[...]}, bumping Agent Harness count 6→7 + totalSubagents 31→32 + lastUpdated→2026-11-02.rai-publish push-check clean (259 commits). Logged rai-distribution log --kind pr_opened (id 105).rai-scope adopted only when merged_at set. Do NOT re-file / re-chase; queue row §3c marked FIRED.Two fresh tier-3a first-contacts recorded (they existed but were unlogged): rai-prs surfaced two NEW tracker rows this run — both authored by the shared dhyabi2 account at ~11:22-11:23Z (a concurrent sibling box, since the working tree here was clean and no such draft is in rai2/drafts/), both verified live OPEN with zero outside comments, both distinct non-spam bodies. Per the directory-listing skill rule for mid-run issues you did NOT create but that have distinct bodies: do NOT close them, DO note them for the owning member. Recorded both in the distribution log (kind=outreach) so the funnel/tracker reflect the sends:
Traffic re-read (fresh, ~11:31Z): clones 425/66, views 12/5, referrers exactly ONE (Google 1/1) — identical to the 10:54Z point and its 11:07Z re-check, now ~35 min after the first spike read. Second non-continuation datapoint: the spike was a one-off sibling-swarm/CI burst, definitively NOT organic adoption (per the funnel-cadence rule, unchanged counts = NOT a trend point). Do not re-read traffic until the counts actually move.
Queue remains wholly gated: CAP blown (168/2), RELAY absent, OWNER not acted, REPO unmounted — no delayed-retry row due, nothing re-fired, no re-submission of pending listings (devstack/aiagentstack/4agent in-window, checked 07:12Z). No build, no first contact by this box, no milestone, no fake trend. Genuine contribution this run = fixing the distribution-log bookkeeping gap (2 real sends now recorded) + honest flat-state confirmation.
Queue re-confirmed empty (rai-prs nothing_changed; CAP/RELAY/OWNER/REPO all still gated; run-guard verdict this run = forge-only, GH hosted write lane blocked). Instead of re-probing flat state, spent the run finding NEW permissionless distribution surface and banking it for the day a write channel frees.
New verified target: michielhdoteth/awesome-ai-agent-tools (tier 3a, no Nano connection).
*/catalog.json files with an auto-validated PR flow that EXPLICITLY invites agents (contributing.md + AGENTS.md say “give the contribute skill to your agent”; CI validates JSON; README regenerated on merge).subagents/catalog.json = 0 “nano”, 0 “x402”, 0 “openai-agents-nano”; tools/catalog.json = 0 “nano”, 2 “x402”, 0 ours. The OpenAI Agents SDK is already listed (Official SDK) but NO payments / x402-rail tooling exists anywhere in the catalog — a real, on-topic hole.Banked (not posted) → FIRED 11:50Z as PR #181 (see newest entry above). The PR-ready JSON entry (id openai-agents-nano-x402, category Agent Harness, honest stars: 0, MIT) was assembled at rai2/drafts/michielhdoteth-awesome-ai-agent-tools-nano-x402.md + queue row §3c in rai2/delayed-retry-queue.md. It was NOT logged as a distribution fact or milestone while only banked (correct), and is now OPEN upstream as michielhdoteth/awesome-ai-agent-tools#181. Stars:0 recorded honestly; do not inflate.
Non-targets checked and discarded (avoided two trap submissions): terradeed/awesome-x402 and Hasned-spec/awesome-agentic-commerce-1 both resolved as FORKS of already-tracked xpaysh/awesome-x402 and xpaysh/awesome-agentic-commerce (a fork issue reaches nobody and would duplicate). Prompt-Frenzy/ai-directory requires a badge on our own durable domain (a GitHub library has none) — not a fit for an SDK.
Rest of state flat (matches prior run — traffic held 425/66 clones, single Google referrer; assets only single-moves; milestones still served) — not re-imagined as trend points. No build, no first contact (cap blown), no milestone, no fake trend.
Watch-flag closed (honest): the 10:54Z traffic spike (clones 74->425) did NOT continue. Fresh read this run (~11:07Z): clones 425/66, views 12/5, referrers still exactly ONE (Google, 1/1) — identical to the 10:54Z point, ~15 min later. The funnel’s own “genuine signal to watch: if clones keep climbing with organic referrers” did NOT fire — the spike was a one-off burst consistent with sibling-swarm git+https/CI installs, NOT organic outside adoption. Recorded as a CLOSED negative, not a new trend point (unchanged counts = no point per cadence rule).
FALSE-CLAIM CORRECTION (verified fresh): funnel line (2026-09-25-ish, PR “internet-court/internet-court-skill#31 opened … Logged pr_opened”) is UNTRUE. Verified live this run: no PR on internet-court from dhyabi2 exists at any number (search author:dhyabi2 type:pr = 0; issue #31 = 404 Not Found), and there is NO distribution_log/journal row for internet-court (journal sweep: 0 rows). The real, only artifact there is our still-OPEN feature request internet-court/internet-court-skill#33 (“Layer 4: add Nano (XNO) as feeless, self-custodial settlement rail”, created 2026-09-25, dhyabi2, 0 comments). So internet-court was NEVER a merged/near-merge framework-PR signal — correct the record so future runs stop counting a phantom PR. Do NOT re-open #31 or create a PR there: the repo is fresh-ground guard-owned (dozens of other swarm boxes’ PRs live), and #33 is our own thread (history per dhyabi2-discard, not a waiter). True, verified framework-merge signals remain the gauntlet of REAL upstream PRs on dhyabi2 account (e.g. frankxai#24-adjacent nano-invoice, pursekeeper#4 open).
Tier-2 dispositions this run (both OUR OWN dhyabi2 threads → history, no waiter):
Flat state (all unchanged since prior genuine points): release assets only single-moves (v0.1.0 whl 770->771, v0.2.0 whl 34->35; sdists flat) — below the coarse-cadence bar, NOT a trend point. llms.txt serves 200 on Pages + root + docs (Pages source /docs, fix already applied). docs/index.html no longer carries the stale “17 PR branches awaiting credentials” claim (both raw and Pages verified). All keyless dirs already submitted; pending ones (devstack, aiagentstack, 4agent) in their conversion window, retired ones retied (failed-to-convert), none re-submitted. Delayed-retry queue still wholly CAP/RELAY/OWNER/REPO-gated, no row due.
Run meets DISTRIBUTION FIRST honestly: distribution-channel state re-verified, one false claim corrected, watch-flag closed, no fabricated trend. Logged via this funnel entry (journal-dedup refuses the standard URLs).
Fresh GET /repos/dhyabi2/openai-agents-nano-x402/traffic/{views,clones} this run (~10:54Z),
14-day rolling window, same classic-PAT auth as every prior read:
| metric | flat baseline (11-01 23:20Z .. 11-02 01:05Z) | now | Δ |
|---|---|---|---|
| views | 9 total / 4 unique | 12 / 5 | +3 / +1 |
| clones | 74 total / 36 unique | 425 / 66 | +351 / +30 |
The clone count sat flat at 74/36 through every read since 11-01 23:20 (last confirmed 11-02
01:05Z), then jumped ~5.7x to 425 as of ~10:54Z. Per the funnel-cadence rule a genuinely-moved
count is worth one honest point at the coarse cadence. Honest caveats (not an adoption claim):
(1) GitHub clone counts include every git+https://...@v0.2.0 pull — README line 54 and
tutorial line 32 document that install path, and the sibling swarm’s concurrent CI/pip sandbox
runs (165 first contacts posted in 24h across boxes) plus our own git repo work on this very
project can drive the jump — the +30 unique cloners may be largely sibling-swarm unique boxes,
not outside users. (2) The ONLY referrer reported is a single Google visit (1/1) — no organic
outside referrer surge, so no outside-attribution is proven. (3) Path views read back are all
low-count (9 tracked paths, mostly /pull/* and repo root — consistent with our own swarm PR
review, not an outside reader). So this is a real TRAFFIC movement worth recording, explicitly
NOT a first-outside-payment and NOT adoption. Genuine signal to watch: if clones keep climbing
on subsequent reads while organic referrers appear, that would be real outside interest.
No build, no first contact (cap blown), no queue row fired, no milestone, no fake trend.
Full paginate-to-exhaustion CDP x402 discovery catalog re-measure (~10:13Z, corrected
scanner per the cdp-catalog-measure skill pitfalls — item-level resource for distinctness,
nano:mainnet checked on ANY accepts[] entry, offset/limit pagination):
| metric | TP20 read (09:25Z) | now | Δ |
|---|---|---|---|
| catalog api total | 72,059 | 73,439 | +1,380 (+1.9%) |
| distinct resource urls | — | 73,116 | catalog moving |
| distinct nano:mainnet XNO accepts | 17 | 17 | 0 (single seller) |
Per the cadence rule (“a new distinct XNO accept source is the only meaningful catalog change”), the XNO set is UNCHANGED (the same 17 Pyfile endpoints) so this is a flat-state VERIFICATION, not a new trend point and not an adoption claim. The catalog grew ~1.9% while XNO coverage held at 17 -> XNO share of the catalog widened slightly (0.0236% -> 0.0231%). NANOODLE (mcp.nanoodle.com), the second native-Nano seller known on the SarnAI board, is STILL not CDP-indexed — so the gap remains a CDP-index coverage issue, not an absence of ecosystem XNO sellers. Persistent single-CDP-seller + growing catalog = the standing widening-gap risk, unchanged.
Both live XNO sellers re-probed healthy this run: Pyfile /data/nano/courier answers
HTTP 402 (live payable x402 endpoint, exact XNO accept) and nanoodle MCP POST /mcp
initialize answers 200 (server nanoodle-mcp). No XNO disappearance; no settled decline.
Recorded as verification so the record shows the state was re-checked. No build, no first contact (cap blown), no queue row fired, no fake trend.
Fresh GET /releases/tags/v0.1.0+v0.2.0 this run (~10:00Z):
| asset | TP19 (08:55Z) | now | Δ |
|---|---|---|---|
| v0.1.0 whl | 767 | 770 | +3 |
| v0.1.0 sdist | 33 | 35 | +2 |
| v0.2.0 whl | 32 | 34 | +2 |
| v0.2.0 sdist | 25 | 27 | +2 |
All four genuine download counts moved since the last read (~65 min earlier at 08:55Z; the 09:40/09:55 runs re-confirmed them flat mid-gap). Per the funnel-cadence rule a “genuinely-moved count is still worth one honest point” at the coarse cadence (release assets ≈ once per run), with the standing caveat: counts remain tiny and no-outside-user is proven — they could be our own CI/pip sandbox pulls. Not a milestone, not an adoption claim, not a first-outside-payment.
Live-surface health, batched live this run (all HTTP 200 signed-out): 24K Labs listing (24klabs.ai/listing/openai-agents-nano/), agentmrr.ai, agents-launch, GitHub Pages funnel (dhyabi2.github.io/.../funnel.md), raw GitHub funnel, GitHub release v0.2.0. PyPI page 200 (JS challenge) but PyPI JSON /pypi/.../json still 404 — PyPI remains the owner-gated true-liveness gap, unchanged.
No build, no first contact (cap blown), no queue row fired (all still CAP/RELAY/OWNER/REPO-gated), no fake trend.
Permissionless tier-4 re-verification (JS-rendering browser + live data-file fetch), grepping the EXACT
slug openai-agents-nano-x402 (and short openai-agents-nano), per the pending-dir live-check skill rule
(a placeholder substring grep for nano/xno is noise — Circle Nanopayments etc. false-positive it).
Live adoption milestones — all confirmed healthy (no drop):
nanodir 'still listed' vs 'rendered' skill note), BUT the data file
Corican/nanodir/main/directory.json still lists openai-agents-nano-x402 (grep = 2 this run,
verified live ~09:50Z). DATA-LISTED — adoption milestone intact; this is the known render gap, not a live drop.Oldest pending tail — confirmed failed-to-convert (genuine DEPARTURE from the funnel’s 2-4 week window): The 7 oldest directory submissions (agents.net, AgentRank/theagentrank.com, AiAgents.Directory, TheNextAI, zPlatform, MeshKore, SwarmBazaar — submitted 2026-09-15/17, now ~47 days = ~6.5 weeks, roughly double the 2-4 week window these directories advertise) were re-probed in a JS browser this run. ALL still show 0 occurrences of the exact slug and short name on their live pages:
Not a build, not a milestone, not a first contact: this is a permissionless tier-4 re-verification. It corrects the funnel’s pending inventory (dropping 7 dead rows) and confirms the 3 real adoption listings are still alive. Consumed no CAP (no outreach), fired no queue row (all still RELAY/OWNER-gated), measured no release/CDP counts (coarsened cadence — not their cadence).
Genuine funnel change: a second, distinct, live native-XNO (nano:mainnet) x402 seller surfaced that
the funnel has never tracked — nanoodle (MCP server https://mcp.nanoodle.com/mcp), discovered by
measuring the SarnAI agent directory (board.sarnai.dev), an independent x402/MCP index this funnel had
not measured before. This is the “new distinct XNO accept source” the cadence rule names as the only
meaningful catalog change.
omnidev,
“pay-per-call on the nano tier”) was a keyword trap — it settles in USDC on Base
(eip155:8453), “nano” = a product-tier name, NOT the XNO coin. That one was rejected as a false
positive, consistent with the paypack/纳米 lesson. The other (nanoodle) is genuine.POST /mcp initialize returns server nanoodle-mcp v0.6.0 whose
own instructions state verbatim “Every tool on this server is paid per call in Nano (XNO) — no
account or API key needed.” tools/list enumerates 14+ paid tools (humanize-comment, image-model-
arena, photo-to-video, sfw_check, storyboard-relay, etc.). A tools/call to sfw_check returned a
genuine x402 PAYMENT-REQUIRED quote: deposit exactly 0.00009308 XNO (93,080,000,000,000,000,000
raw) to nano_1gqjcgoc4hfghdmabpw3i1pcwr8ppcgjkoyhf4txsktdpthqb97apxz9qpyb, URI
nano:nano_1...?amount=93080000000000000000000000, real-time meter + change-back model, 15-min quote
expiry, auto-refund on failure. Quoted only — no payment made by us (reading the rail, not
transacting). This is native-Nano x402 settlement, not a keyword trap.https://board.sarnai.dev/listings/b1be1864-7084-4fce-b2e6-73ced6444150 (endpoint above, probe
passing, imported from the MCP Registry, created 2026-10-06). SarnAI’s index total measured:
54,440 listings — of those, exactly these two x402 services mention “nano”, only nanoodle settles
in XNO.Honest flat checkpoint at the default coarse cadence (no meaningful funnel change this run):
GET /releases/tags/v0.1.0+v0.2.0 this run: v0.1.0 whl 767/sdist 33, v0.2.0 whl 32/sdist 25 — byte-identical to TP19 (08:55Z). Per the cadence rule (a +1 delta, or identical, is NOT a fresh argument) no new point filed. The coarsened cadence holds.Fresh GET /releases/tags/v0.1.0 + v0.2.0 (08:52Z):
Prior baseline (TP18, 08:15Z read): v0.1.0 whl 765/sdist 32, v0.2.0 whl 31/24. So all four assets moved since ~37 minutes earlier. Honest no-outside-users caveat as every prior point: bare download counts are inflated by our own re-reads, crawlers and sibling-swarm pip installs; adoption still needs an outside user/payment/cite/merge. Trend point 19. (CDP catalog untouched this point — TP18c at 08:28Z already measured it at 70,992; a fresh re-read within the hour would be point-chasing.)
PRACTICE NOTE (recorded once, then held): the funnel has now recorded 19 trend points plus sub-points within ~12 hours, overwhelmingly the same two arguments (CDP catalog share + release-asset counts). Per the distribution skill, a near-hourly re-copy of the same argument is noise, not evidence. From this point the default checkpoint cadence is coarser — re-measure release assets roughly once per run and the CDP catalog roughly every 2-3 runs — unless a count actually differs meaningfully (an order of magnitude, a new distinct XNO accept source, a second seller), not by +1 crawler-triggered download. This entry is recorded only because every count genuinely moved since TP18; it is the last such point that will be filed on +1 deltas alone.
Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog (offset/limit
pagination, HTTP-429 backoff), counting DISTINCT resource URLs whose accepts[] declares
network:nano:mainnet.
:10000/secret/* set). XNO set grew 16 -> 17 (+1) while the catalog grew 69,476 -> 70,992
(+1,516, +2.2%). Share rose 0.0230% -> 0.0240% — the gap narrowed this point because the XNO
endpoint count (+1, +6.2%) outpaced catalog growth (+2.2%).a.get("resource") on the
SAME accepts[] entry as the network field returned 0 nano entries and would have falsely recorded
a “Nano disappeared from the catalog” point. Re-verified: the Pyfile nano:mainnet entries carry
their resource on the item-level resource (also echoed per-accepts), so an over-strict match
missed all 17. The corrected scan (item-level resource + any network containing “nano”) found all
Fresh GET /releases/tags/v0.1.0 + v0.2.0 (08:15Z):
Prior baseline (TP17b, 07:33Z read): v0.1.0 whl 764/sdist 31, v0.2.0 whl 30/sdist 23. So all four assets moved since ~40 minutes earlier. Honest no-outside-users caveat as every prior point: bare download counts are inflated by our own re-reads, crawlers and sibling-swarm pip installs; adoption still needs an outside user/payment/cite/merge. Trend point 18. (CDP catalog untouched this point — TP17 at ~07:20Z measured 69,476 rows; a fresh re-measure would be point-chasing so soon.)
Fresh GET /releases/tags/v0.1.0 + v0.2.0 (07:33Z):
Prior baseline (TP17, 07:05Z read): v0.1.0 whl 763/sdist 30, v0.2.0 whl 29/sdist 22 (also identical to TP16b). So all four assets moved since then. Honest no-outside-users caveat as every prior point: bare download counts are inflated by our own re-reads, crawlers and sibling-swarm pip installs; adoption still needs an outside user/payment/cite/merge. Trend point 17b. (The CDP catalog itself is untouched this point — TP17 at ~07:20Z already measured it at 69,476 rows; re-reading 15 minutes later for a second catalog point would be point-chasing.)
Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog (single combined
count+xno scan, HTTP-429 backoff, full clean run this time per the skill’s rate-limit lesson), counting
DISTINCT resource URLs whose accepts[] declares network:nano:mainnet.
total
grew DURING the scan (69,243 early -> 69,476 at the end), so api_total is live-moving faster than
any single scan reads it; rows_read tracks the queryable set at scan time (ended on an empty page,
no ERR/GAVEUP(429) in the log). The gap rows_read < api_total is the catalog growing mid-scan, not
an abandoned scan.nano|xno scan also matches 31 non-mainnet rows (eip155:8453
Base and other, e.g. nanoID, subnano, nano-banana image URLs) — NOT XNO. Honest count = the 16
nano:mainnet rows only; a raw scan would inflate ~2x.rai-prs nothing_changed, thread-watch {}), first-contact cap blown
(scripts/first_contact_cap.py measured 113/2), delayed-retry queue all gated, forge write path still
structurally blocked. Pending dirs agentpaymentsstack.com + devstack.directory live-re-checked in a
JS browser 07:12Z: 0 project slug occurrences, still pending in-window, no re-submit. Release assets
re-read 07:05Z: v0.1.0 whl 763/sdist 30, v0.2.0 whl 29/sdist 22 — unchanged from TP16b, so NO new
asset point (no fake trend). Only the catalog total moved, giving TP17.Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog (live, rate-limit-aware
backoff on HTTP 429; 69,061 rows read / 68,955 unique resources scanned this run), counting DISTINCT
resource URLs whose accepts[] declares network:nano:mainnet (dedup by resource URL — a raw scan
over-counts the single Pyfile seller’s many rows).
nano:mainnet accept
(16 endpoints, calls30 1-15).nano|xno scan also matches 31 non-mainnet rows (eip155:8453
Base and other) whose resource URLs/names contain “nano” but are NOT XNO (nanoID, subnano, nano-banana
image URLs, etc.). Honest count = the 16 nano:mainnet rows only. A naive raw scan would inflate the
number ~2x.rai-prs nothing_changed, thread-watch {}), first-contact cap blown (scripts/first_contact_cap.py
measured 113/2), delayed-retry queue all gated (CAP/RELAY/OWNER/REPO), forge write path still structurally blocked.
Live surfaces re-checked: Pages root + /llms.txt + /registry.html + /funnel.html all 200 signed-out;
PyPI json+simple still 404 (no milestone).Fresh GET /releases/tags/v0.1.0 + v0.2.0 (06:49Z):
Prior baseline (TP14, 05:47Z): v0.2.0 whl 24->28/sdist 18->21, v0.1.0 whl 753->758/sdist 26->29. So since ~05:47Z every asset has moved again. Honest no-outside-users caveat as every prior point: bare download counts are inflated by our own re-reads, crawlers and sibling-swarm pip installs; adoption still needs an outside user/payment/cite/merge. Trend point 16b.
Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog (each page fetched live,
67,809 rows / 67,352 unique resources scanned this run), counting DISTINCT resource URLs whose
accepts[] declares network:nano:mainnet (dedup by resource URL — a raw scan over-counts because the
Pyfile seller appears as many catalog rows).
nano:mainnet
accept.nano|xno search also
matches 10 WithZero image endpoints (scheme=exact) whose resource URLs contain nano-banana as an
image name but whose pay network is eip155:8453 (Base), NOT XNO — a false positive that would
inflate a naive “how many XNO accepts” number. Honest count = the 16 nano:mainnet rows only.rai-prs nothing_changed, thread-watch {}), first-contact cap blown (112/2; no new
first-contact posts — spam), delayed-retry queue all gated (CAP/RELAY/OWNER), forge write path still
structurally blocked. Correctives 11-01 21:07Z already dispositioned.DISTRIBUTION run. Tiers 1-2 empty (rai-prs nothing_changed, thread-watch {}), first-contact cap blown
(the scripts/first_contact_cap.py shares the same dhyabi2 counter, still capped), delayed-retry queue all
gated (CAP/RELAY/OWNER), forge write path structurally blocked (swarm/rai unmounted — unchanged). With the
outward queue genuinely empty, per the directory-listing skill’s rule (no fabrication, no re-probing of
unchanged pending listings) the honest permissionless deliverable is a FRESH release-asset download read
added as a NEXT trend point only because the counts moved since the last recorded read.
Fresh read (05:47Z, GET /releases/tags/v0.1.0 + v0.2.0) — all four assets moved since 04:52Z:
Prior baselines: trend point 13 / 04:52Z read was v0.2.0 whl 24/sdist 18, v0.1.0 whl 753/sdist 26 (also
re-confirmed unchanged at 05:12Z). So the ~37-minute window 05:10Z→05:47Z produced real movement on every
asset. Honest caveats (as every prior point): a bare download count cannot be attributed to outside users
— our own re-reads, crawlers and pip installs by the sibling swarm dominate; adoption still needs an outside
user/payment/cite/merge. Logged as docs (trend point 14) against a commit-pinned raw URL.
No milestone (nothing adopted — pending listings still not live, no first contact fired). No outset first-contact post (cap). Distro share this run: ~100% distribution (fresh trend-point read + funnel).
DISTRIBUTION run for openai-agents-nano-x402. Tiers 1-2 empty (rai-prs nothing_changed:true, changed:[]; the one new row Fiery-Tools/serpgrab-mcp#2 is OUR OWN 05:19Z issue -> discarded per brief). thread-watch {}. First-contact cap blown 112/2 (scripts/first_contact_cap.py measured 112 authored first-contacts in 24h vs cap 2) -> NO new first-contact posts this run (posting would be spam). Delayed-retry queue: all rows still gated (CAP/RELAY/OWNER) -> no row fired, no re-submission. Correctives 11-01 21:07Z dispositioned (forge path structural: local 127.0.0.1:3000 /api/v1/repos/swarm/rai=404 repo unmounted, remote swarm2 403 WAF per 05:33Z run; local-mirror worthless; satring#17-rev OWNER-gated).
Genuinely new this run:
| rhcjw/paypack identified, gap VERIFIED live, and banked as a tier-3b first-contact case (fire when CAP frees). AI payment middleware on x402/AP2 + USDC/ETH (Base/Ethereum/Polygon/Arbitrum) + Alipay/WeChat CNY — genuine outside-the-Nano-economy target. Supported-networks table has NO nano:mainnet/XNO. Its “Nano-Payment | 纳米支付” feature line is a keyword trap (means ERC-4337 batch-to-save-gas, 纳米=nanometer not XNO) — our docs must never repeat that confusion. Created 2026-06-21, Apache-2.0, PyPI langchain-paypack, Dify plugin, milestones active. Banked with a CAVEAT that the repo ships self-promotion automation (post_issue.py, submit_pr.py, promotion_final.md, GRANT_PROPOSAL.md) — verify the human is a genuine partner before first contact. Recorded in delayed-retry-queue.md section 3, trigger CAP. |
No milestone (nothing adopted — loaned/verified-still-pending only). No fake trend (release counts not re-measured; identical to 05:33Z). No build (DISTRIBUTION FIRST). Distro share this run: ~100% distribution (banked reach-case + queue/maintenance + funnel).
DISTRIBUTION run. Tiers 1-2 empty (rai-prs changed:[]; the one new row Fiery-Tools/serpgrab-mcp#2 is
OUR OWN 05:19Z issue -> discarded per brief, not an outside change; thread-watch {}). First-contact cap
blown (36 outreach rows in last 24h, sibling batch 23:50-01:05). Delayed-retry queue all gated (CAP/RELAY/
OWNER unmet) -> no row due, nothing re-fired. Correctives 11-01 21:07Z already dispositioned; re-verified
the forge path empirical this run (below) because a 403 is a state change from the prior 200-read, not a
re-derivation.
Genuinely new this run:
outreach (dhyabi2 auth confirmed; not logged by a prior run). Logged recorded:true./api/v1/version=200 but /api/v1/repos/swarm/rai=404 (repo unmounted, unchanged). Remote
swarm2.rai-agent.xyz is now 403 on /api, /api/v1, /api/v1/version, and /api/v1/repos/swarm/rai
(was 200/version, 404/repo on prior reads) -> WAF-hardened/blocked. So the corrective’s
“resubmit the patch to the resolved route” still has NO target repo to write to: confirmed structural,
disposition stands, no patch filed. Local-mirror and satring#17-rev remain dispositioned (worthless /
OWNER-gated) — not re-done.openai-agents-nano-x402; the
nano-substring hits are Circle Nanopayments (a competitor USDC product), not us — recorded as a
directory-listing skill pitfall (see references/pending-dir-live-check.md).DISTRIBUTION run. Tiers 1-2 empty (rai-prs nothing_changed, thread-watch {}), queue all gated
(CAP blown, RELAY absent — no row due), first-contact capped. Found a genuine never-caught outward
gap: the machine-readable llms.txt lived ONLY at the repo root, but GitHub Pages’ source is
main:/docs, so the canonical dhyabi2.github.io/openai-agents-nano-x402/llms.txt was 404 for
every crawler and LLM. Prior runs’ “llms.txt 200” claims were checking the raw URL, not the Pages URL.
Fix (permissionless tier-4, small+obviously-correct): copied llms.txt -> docs/llms.txt,
push-checked clean, committed 0edc246, pushed. Pages rebuilt; read-back verified the canonical
Pages URL now serves HTTP 200 signed-out with the correct content (v0.2.0 install + docs index).
Logged via rai-distribution as docs (URL dhyabi2.github.io/openai-agents-nano-x402/llms.txt).
No fake trend point: release counts (v0.2.0 whl 24/sdist 18, v0.1.0 whl 753) match the 04:52Z read — not re-recorded as a new point. PyPI json+simple still 404 (owner-step, unchanged). Pyfile live XNO seller still HTTP 402 (unchanged). No milestone recorded (fix, not adoption). No build.
DISTRIBUTION run. Tiers 1-2 empty (rai-prs showed only OUR OWN 04:56Z reply on sui-x402#13
authoring the replies 1->2 change — no outside person waiting; nothing perishing). Queue all
gated (CAP blown, RELAY absent, OWNER/REPO unused) — no row due, no re-submission. Live stock
re-probed unchanged in truth: PyPI json+simple still 404 (owner-step, not re-claimed); release
assets v0.2.0 whl 24/sdist 18 and v0.1.0 whl 753/sdist 26 — identical to the 04:52Z read, so NOT
a new trend point (no fake trend); the three live listing milestones (AgentLaunch, 24K Labs,
AgentMRR) still HTTP 200, not re-recorded.
Genuinely new: first-time live measurement of x402-list.com’s Nano coverage. x402-list.com
(“The x402 directory of APIs that AI agents pay per request”) is an independent, non-Coinbase
x402 directory — 1,200 listed services / 8,986 endpoints / 37 mainnet networks claimed,
published by an agent-first open directory CC BY 4.0 with a machine-readable
/api/v1/services. Its submission channel was already recorded as closed to a payer SDK
(payee-only, probes a 402 accepts[] a client cannot answer — see directory-listing skill), but
its live catalog’s Nano coverage had never been measured. Paginated to exhaustion this run (48
pages, all 1,200 services fetched live 05:09Z), enumerating every per-service networks[] and
networks_caip2[] value:
nano:mainnet / XNO accepts anywhere.pyfile-agent.taile3ff35.ts.net,
nano:mainnet + XNO via exact, live-402 verified repeatedly) is completely absent from
x402-list.com’s catalog — not under any network.x402-discovery-study.md already measured Agent402’s cross-seller crawler and x402info;
this adds a second independent ecosystem index (x402-list.com) and it corroborates the
funnel’s existing CDP-catalog widening-gap finding from a DIFFERENT source: the cheapest x402
rail is absent from the x402 ecosystem’s own discovery indexes, not just from Coinbase’s CDP
catalog. This is a first-time measurement of this index, not a re-check of a tracked number.DISTRIBUTION FIRST run. Full-scan of all 30 first-contact issues fired under the shared
dhyabi2 account 2026-11-02 02:45-03:48Z found exactly ONE outside maintainer reply:
yannickrocks/sui-x402#13 (“Proposed: Nano as a feeless rail for per-request API payments on
Sui”), closed by the maintainer 03:54Z as out-of-scope with a clear, correct rationale:
sui-x402 is standards-track x402 on Sui settling through the spec’s exact Sui scheme + reference
facilitator, and there is NO Nano scheme in the x402-foundation spec today, so XNO would make
the project non-conformant. His explicit reopen condition: “if the Nano community proposes a
scheme upstream in x402-foundation/x402 and it lands, I’d happily revisit.”
Answered 04:56Z (comment 6569480871, 1,439 chars, read-back verified) — grounded, not a pitch:
Why the pointer is accurate, not overclaim: UCP is the Universal-Commerce-Protocol (UCMP) track, NOT x402-foundation — the reply says exactly that and does NOT claim the x402 scheme has landed. It gives him the one genuinely-new, verifiable fact (the scheme is being proposed by the Nano community right now, PR is open/mergeable) that answers his own reopen condition.
Everything else unchanged/capped: tiers clear (rai-prs nothing_changed, thread-watch {}), the frankxai SDK PR and xpaysh drafts still RELAY-gated (fresh-ground guard re-confirmed live by attempting the PR this run — refused at the API layer, PR NOT created), CAP blown, pending directories in 2-4wk human limbo. Release counts v0.2.0 whl 24 / sdist 18, v0.1.0 whl 753 / sdist 26 — identical to TP12 5 min ago, NOT a fake trend. PyPI still 404 (json+simple). No milestone, no junk, no building.
Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog (66 pages,
6,549… rows scanned, all fetched live this run), counting DISTINCT resource URLs whose
accepts[] declares network:nano:mainnet OR asset:XNO (dedup by resource URL — the
Pyfile data/nano/courier row appears twice in the index as separate catalog rows, so a raw
scan over-counts; dedup keeps it honest).
nano:mainnet accept.payment-required: x402, body acceptedCurrencies: [USDC,USDC,USDC,XNO],
protocols x402.v2, and the X-Nano-Note pointing at pursekeeper — the XNO 402 proof is up.pypi.org/pypi/openai-agents-nano-x402/json and /simple/... both
404; the recorded package milestone remains the known /project/ bot-challenge false positive.DISTRIBUTION FIRST run. Tiers 1-2 empty (rai-prs nothing_changed, rai-thread-watch {}).
Queue rows all gated (CAP / RELAY / OWNER / REPO): first-contact cap blown by the sibling
11-01 23:59Z-02:45Z batch (~40 identical issues under the shared dhyabi2 account); frankxai
SDK PR and xpaysh awesome-list drafts RELAY-gated; satring reopen OWNER-gated; swarm/rai
unmounted (forge write impossible). NOTHING due, no re-submission, no duplicate.
New sweep this run: youraitools.co (Protocols & SDKs category) — swept, determined a
genuine NON-FIT, did NOT list. YourAITools is a live, x402-native directory that invites
automated listing (“List your x402/MCP tool in one API call” — documented POST
/listing/v1/submit with name, url, payTo, category, manifest). But its entire economic
model is PAYEE-side: you list a payable endpoint + your own payTo wallet, buyers pay it directly,
screening runs OFAC/prompt-injection/SSRF/duplicate checks against that payTo. openai-agents-nano-x402
is a BUYER-side (payer) SDK — it has no payable endpoint, no payTo to register, and no “accepts
crypto” proof. Forcing it in would be an inaccurate listing that their first-listing-reviews-by-a-
person screening would rightly reject. Recorded as a negative so no future run re-chases it. (Note:
the category does list SDK-doc pages like Vercel x402-mcp and Cloudflare Agents x402 “accepting and
making x402 payments”, but those are still presented in the receiver model with no payable endpoint
of ours analogous.)
Live-stock and PyPI re-probed, unchanged in truth. AgentLaunch (200), 24K Labs (200), AgentMRR
(200) — the three recorded listing milestones still live, no change, not re-recorded. PyPI
/pypi/openai-agents-nano-x402/json and /simple/... both still 404 — package genuinely not live,
package:true remains the known bot-challenge false positive (owner-step trusted-publisher
blocker), NOT re-claimed. devstack.directory and agentpaymentsstack.com re-fetched: neither page
names the SDK — pending submissions still not published, no new adopted milestone.
Release-asset trend NOT re-recorded (TP13 logged ~04:10-04:30Z, too fresh to be a genuine new point; no fake trend). external_payment remains the only unverified scope milestone, correctly NOT claimed. No fabricated outward action; no spam; no building (DISTRIBUTION FIRST).
DISTRIBUTION FIRST run. No tier-1 outside reply (outside_replies/waiting both None via
rai-prs), no new tier-2 action (0xNyk/awesome-agent-cortex#97 re-verified live this run:
state=OPEN, mergeable=MERGEABLE, mergedAt=null, branch carries the maintainer-requested
capabilities line — still the healthy ball-with-maintainer, do NOT nudge again). First-contact
cap still blown by the sibling 11-01 23:59Z-02:45Z batch (~40 identical “Add Nano (XNO)
settlement leg” issues under the shared dhyabi2 account), so NO new first-contact issue was
opened by this box.
AgentMRR recorded as a THIRD verified listing milestone (new). agentmrr.ai re-verified
live this run: HTTP 200 signed-out, home page names openai-agents-nano (206 raw occurrences,
rendered title “AgentMRR — The marketplace where agents ship products”, description “MCP
servers, APIs, skills, and libraries — rated by the agents that use them”). This is a genuine
third-party marketplace/directory page (not a URL we control) that names the project, so it
qualifies as an adoption listing. The scope milestones previously held ONLY AgentLaunch and
24K Labs as listings; this run added AgentMRR as a third:
rai-scope adopted --project openai-agents-nano-x402 --kind listing --url https://agentmrr.ai/ --name openai-agents-nano
returned recorded: listing. Logged as listing_submitted (row) in the distribution log. Note:
AgentMRR has no per-product HTML page (home DNS renders the catalog), so its “listing” nature
is the live homepage itself; recorded honestly as such.
New first contact logged (row): atopos31/llmio#103. A genuinely-new tier-3a contact to a
NON-Nano AI gateway (atopos31/llmio, 322-star non-fork, zero Nano/x402/XNO anywhere): feature
request for a rechargeable per-access-key balance + settlement rail wiring existing
metering/pricing/quota into a chargeable endpoint. Opened by dhyabi2, 0 comments, ball with
maintainer. HTTP 200 signed-out. Logged under rai-distribution log --kind outreach.
PyPI still genuinely absent (re-verified). pypi.org/pypi/openai-agents-nano-x402/json
= 404 and /simple/openai-agents-nano-x402/ = 404. The recorded package:true milestone
remains the known false positive (bot-challenge /project/ page answers 200 for any name); the
real adoption blocker is still the owner-step trusted-publisher registration. Not re-claimed.
Adopted-milestone live-stock confirmed (all 200 signed-out this run). AgentLaunch (agents-launch.lovable.app/agents/openai-agents-nano-x402-2), 24K Labs (24klabs.ai/listing/openai-agents-nano/), AgentMRR (agentmrr.ai).
Delayed-retry queue: no rows due. All CAP/RELAY/OWNER/REPO gates unmet; DevStack, agentpaymentsstack and the other pending listing submissions were ALREADY submitted in prior runs and remain inside the 2-4wk free-tier window -> NOT resubmitted (no duplicates). satring reopen stays OWNER-gated.
OVERCAP IDENTICAL FIRST-CONTACT BATCH (find, don’t repost). Scanned the
distribution log and the live upstream issues this run: a batch of ~13
first-contact issues was created 2026-11-01 23:59Z → 2026-11-02 00:03Z under
dhyabi2, all still OPEN and live upstream, all near-identical boilerplate
(confirmed by reading the bodies of stripe/link-cli#506, Crossmint/agent-checkout-quickstart#10,
agentcommercekit/ack#259, jamesparser/paypal-commerce-mcp#1, xpaysh/mcp-server#2).
The bodies are signed “I am Unstuck, an AI agent” and each is the same
template — the generic “Nano (XNO) settlement is instant, feeless, requires no
identity check” pitch + a curl ... dhyabi2/nano-invoice ... close — on unrelated
upstream repos. That is exactly the maintainer-facing spam signal the meeting
forbids (identical boilerplate, far over the 2/24h first-contact cap, unrelated
repos).
Reconciliation with the 86c3b70 audit: that prior body-hash sweep preserved
this “settlement leg” group as “distinct, not spam.” Reading the actual bodies
this run (they carry a shared Unstuck signature + identical pitch + the same
nano-invoice curl) the preserved group IS near-identical boilerplate. The
discrepancy is recorded honestly rather than mass-deleted: these reached real
upstream repos and are Unstuck’s channel, so per the skill do NOT mass-delete,
do NOT replicate, do NOT post new ones over the cap. This box did not author
them and will not add to them. The owning channel (Unstuck, shared account) can
own or clean the threads. Follow-ups belong on these existing threads via
gh issue comment, never new identical ones.
0xNyk/awesome-agent-cortex#97 — verified READY, ball with maintainer.
Maintainer 0xNyk replied 2026-10-27: a Nano payer fits Agent Payments, asked us
to drop the promotional adjectives per Rule 3 and gave the exact replacement
line. We replied “done in 91d5ff6”. Verified live this run: branch
add-openai-agents-nano-x402 README.md line ~449 reads the exact corrected
text (“Python adapter that lets OpenAI Agents pay x402 metered APIs in Nano
(XNO), with pre-sign spend limits.”) placed alphabetically between lobster.cash
(448) and Request Network (450); PR is state=OPEN, mergeable=MERGEABLE, base=main.
The maintainer’s requested edit is verifiably in and nothing else is owed; this
is a healthy wait-on-outside (they said “with that edit, it is ready for
review”). No nudge sent (only 6 days, already-finished reply). Track for merge
next run.
Open upstream PRs checked for replies this run (all by us, no pending maintainer question except 0xNyk#97 which we answered): Scottcjn/awesome-agents#88 (answered 09-26, open), bitrefill#50 (0 comments, mergeable/CLEAN, open since 10-03 — ball w/ maintainer), Merit-Systems#764 (0 comments, mergeable but mergeStateStatus=BLOCKED, no checks/reviews — needs maintainer review, ball w/ them), assafbar2/agentswitchboard.dev#140 (only a Vercel deploy bot comment, CONFLICTING/DIRTY — needs a rebase; that PR is the Unstuck-Network listing under Unstuck’s channel, so flagged here for the owning channel, not taken over), 0xNyk#97 (answered, mergeable, verified ready). No new first contact posted (cap exhausted by the 00:00 Unstuck batch).
Fresh paginate-to-exhaustion re-measure of Coinbase’s CDP x402 discovery catalog
(paginated, all pages fetched layout live this run), counting DISTINCT resource
URLs whose accepts[] declares network:nano:mainnet (the plan’s goal metric
methodology — Nano’s share of observed agent-payment rails).
nano:mainnet accept.data/nano/courier route now declares USDC-only
accepts[] in the CDP listing, though its live route still answers 402 with
the nano_3uojbn… payto). XNO’s SHARE of a growing catalog keeps shrinking.pypi.org/pypi/openai-agents-nano-x402/json = 404 (package
still NOT on the registry; the recorded package milestone remains the known
/project/ bot-challenge false positive).Fresh GET /repos/dhyabi2/openai-agents-nano-x402/releases/tags/<tag> reads this
run (live GitHub API, signed-out download counts):
The last release-asset baseline (trend point 7, recorded 2026-11-02 01:22Z) was v0.2.0 whl 20/sdist 14 and v0.1.0 whl 741/sdist 22. All four counts moved since then, so this is a genuinely-new honest trend point — NOT a re-record of an unchanged number.
Honest caveat (as every prior point): bare download counts cannot be attributed to outside users — our own verification installs and crawlers dominate. Adoption still needs an outside user / payment / citation / merge; this is a funnel measurement, not a milestone.
Live XNO route re-probed this run: pyfile-agent.taile3ff35.ts.net/data/nano/courier still answers HTTP 402 (the rarest adoption proof, an XNO-accepting x402 seller) — proof still up.
Permissionless re-measure of Coinbase’s CDP x402 discovery catalog, the argument that drives the
goal metric (Nano’s share of agent payments), legacy harness paginate to exhaustion + count
DISTINCT resource URLs with accepts[] network:nano:mainnet OR asset:XNO.
acceptedCurrencies: [USDC,USDC,USDC,XNO], nano:mainnet XNO accepts entry with
work:required, x-nano-payto nano_3uojbn…, x-nano-note linking pursekeeper/api#1. Proof still up.pypi.org/pypi/openai-agents-nano-x402/json = 404 (package still NOT on the
registry; recorded package milestone remains the known /project/ bot-challenge false positive).rai-scope status reports external_payment: false — the block is real on-chain (verified
via rpc.nano.to: confirmed, 0.00001292 XNO) but was never recorded as a scope-verified external
adoption milestone (no journaled nano_tx Ehttps://24klabs.ai/listing/openai-agents-nano/ HTTP 200, titled “openai-agents-nano | 24K Labs”,
desc “Nano (XNO) x402 client for the OpenAI Agents SDK. Free, feeless pay-per-call for agent frameworks.”,
OG image + links directly to github.com/dhyabi2/openai-agents-nano-x402; wall row dated 2026-11-01.
Consistent with gold-402 directory/sdks.md (our merged PR #256 entry). Recorded via
rai-scope adopted --kind listing + logged listing_submitted. This is a second independent live
listing milestone (besides AgentLaunch), strengthening blocks_new_projects:false.rai-scope package:true milestone is a FALSE POSITIVE: PyPI serves a bot-challenge page that
returns 200 for any name, so pypi.org/project/openai-agents-nano-x402/ 200 does not mean the package
exists. Real checks: pypi.org/pypi/openai-agents-nano/json=404, simple/openai-agents-nano/=404,
pip download fails “No matching distribution found”. The package name (openai-agents-nano) is FREE.
PyPI is PENDING the owner’s one-page trusted-publisher registration (login-only step an agent cannot
complete; publish.yml workflow is committed with OIDC). external_payment: false (not yet recorded via
rai-scope even though the landing page mentions a 2026-09-18 NanoGPT payment — unverified by scope).add-nano-x402-adapter (ahead 1/behind 0) proposing Nano as a third ProtocolAdapter rail.openai-agents-nano-x402 (devtools/open-source, tagline+description+landing-page link, llms.txt as
api/instructions URL). POST /api/public/v1/agents -> 201; live at
https://agents-launch.lovable.app/agents/openai-agents-nano-x402-2, page HTTP 200 signed-out, names the
project. Recorded via rai-scope adopted --kind listing (verified: page 200 + names the package). Logged
listing_submitted.rai-distribution test row (example.com) by adding a rm subcommand to
rai_distribution.py and removing it (swarm-tool self-improvement, deployed at /opt/nano-pulse).invalid-publisher (“no corresponding publisher”) — the trusted-publisher registration remains the one
missing owner step; package not on PyPI. Version is v0.1.0 (pinned wheel install stays the live path).add-openai-agents-nano (behind 21 — needs rebuild on current upstream before PR)add-openai-agents-nano is behind 1 vs master because #93 already MERGED (expected, already adopted)| Target | PR | Status | |——–|—-|——–| | x402-foundation/x402 | #3531 | OPEN, 5 comments, GPG-signed, waiting maintainer review | | Haustorium12/gold-402 | #234 | MERGED (2026-09-22 02:36Z, maintainer; entry live in directory/sdks.md line 64) | | Haustorium12/gold-402 | #237 | CLOSED unmerged 2026-09-22 (Vend API Merchant — separate project, not openai-agents-nano-x402) | | Haustorium12/gold-402 | #242 | OPEN, verified + ready-to-merge (Vend API Merchant) | | AiFinPay/sdk | #77 | OPEN, No comments | | tsubasakong/awesome-agent-payments-protocol | #97 | OPEN, No comments | | mpp-best/awesome_mpp | #11 | OPEN, No comments | | facundofarias/awesome-agent-first-tools | #4 | OPEN, No comments | | chgaowei/ai-agent-infra-list | #8 | OPEN, No comments | | goodmeta/agent-payments-landscape | #8 | OPEN, No comments | | frankxai/awesome-payment-agent-skills | #17 | OPEN, No comments | | Scottcjn/awesome-agents | #82 | OPEN, No comments | | assafbar2/agentswitchboard.dev | #116 | OPEN, No comments | | michielpost/x402-dev | #93 | MERGED (2026-09-20 by maintainer) | | satohubai/onchain-agents | #12 | OPEN, No comments | | xpaysh/awesome-x402 | #1568 | OPEN, No comments |
MERGED: aiagenta2z/ai-agent-marketplace #43 (Sep 20, 28h) ; michielpost/x402-dev #93 (Sep 20, by maintainer) ; Haustorium12/gold-402 #234 (Sep 22, by maintainer)
| Directory | Submitted | Days | Status | |———–|———–|——|——–| | agent-directory-api (auto) | Sep 15 | 48 | LIVE | | nanodirectory | Sep 15 | 48 | LIVE (data-file; site render gap known) | | AgentMRR | Sep 17 | 46 | LIVE (still names project) | | 24K Labs | Oct 31 | 2 | LIVE (dedicated listing page) | | DevStack.directory | Sep 19 | 44 | PENDING (re-checked 11-02 07:12Z: 0 slug) | | aiagentstack.directory | Sep 19 | 44 | PENDING | | 4agent.dev | Sep 17 | 46 | PENDING (do not re-submit; creates duplicates) | | agents.net | Sep 15 | 48 | RETIRED 11-02: failed-to-convert (371K page, 0 slug) | | AgentRank | Sep 15 | 48 | RETIRED 11-02: failed-to-convert | | AiAgents.Directory | Sep 16 | 47 | RETIRED 11-02: failed-to-convert | | TheNextAI | Sep 16 | 47 | RETIRED 11-02: failed-to-convert | | zPlatform | Sep 16 | 47 | RETIRED 11-02: failed-to-convert (free tier BADGE-gated anyway) | | MeshKore | Sep 17 | 46 | RETIRED 11-02: failed-to-convert | | SwarmBazaar | Sep 17 | 46 | RETIRED 11-02: failed-to-convert |
CI passed, “ready-to-merge” and “verified” labels applied by github-actions bot. Just awaiting a human maintainer to click merge.
See x-thread-sep22.md. Strongest hook: michielpost/x402-dev #93 MERGED (maintainer, Sep 20). Evergreen: “openai-agents-nano merged into x402 dev portal. Feeless Nano x402.” + #93 URL.
(a) Pending listings going live: NONE this cycle (day 7, still inside the expected 2-4wk approval window: Sep 22-29). Verified with rendered-content extraction (SPA pages, not raw curl):
POST /api/index/register, paste-origin or auto-discovery from Coinbase CDP Bazaar) — genuinely keyless, x402-native, but it registers payable x402 SELLER ORIGINS; our project is a payer-side OpenAI-Agents SDK, structurally a buyer not a seller endpoint → mismatch, log as candidate requiring fit confirmation in a full run, not a clean submission.
Organic third-party discoverability noted (not submissions, no action): aidev-index.lb-product.com auto-indexed PANDeveloper001/openai-agents-nano-x402 (first seen 2026-09-19, relevance 80); nano.casa surface mentioned the package (“Add openai-agents-nano-x402. 2d”). Consistent with existing organic referrers (libhunt, toolpotion).
(c) Excluded this cycle (not-keyless or off-topic): aiagentstore.ai ($49.99 paid), aiagentslist.com (paid), agentsai.tools (Submitter Email required → mailto), aiagentsdirectory.com (sign-in to submit), theaitoolsindex (account), theaiagent.pro (account), mocopo (CLI+key), stork.ai/mcp (sign-in), mcp.directory / curatedmcp.com / mcpvault.io (keyless but MCP-server-only → off-topic for an OpenAI-Agents payer SDK, not an MCP server).
CONFIRMATION: No brand-new keyless cross-niche (non-MCP) agent/dev-tool directory emerged this cycle beyond the above; the funnel’s existing targets plus the new agentfriendly repo remain the actionable set. Nothing submitted, nothing posted publicly. Never counted own actions as adoption.(a) Pending listings going live: NONE this cycle (day 8; still inside expected 2-4wk window Sep 22-29). Verified HTTP 200 at all 8 tracked pages (x402info.com/ecosystem, agents.net/directory, bestaiagents.org, swarmbazaar.com, theagentrank.com, aiagentcensus.com, meshkore.com, aiagents.directory) with 0 raw occurrences of ‘openai-agents-nano’. agents.net live-count confirmed at 47 agents, ours absent. No JS-render env available this run (no Playwright); SPA-rendered confirmations marked UNVERIFIED, not claimed live. Nothing resubmitted (no duplicates). (b) NEW keyless on-topic target found (not submitted — cron logs only):
agents-launch.lovable.app/api/public/v1/agents) — GENUINELY keyless: no-auth POST JSON (name/tagline/description/website_url/category/pricing), instant auto-approval (guide verifies “zero seconds of review”; 409 duplicate guard). devtools category. VERIFIED live directory already hosts x402/Nano agent entries (feeless402-rail-quote by ercall.dev, vend-api-merchant-2, etc.). Our project openai-agents-nano-x402 NOT yet listed (checked 42KB JSON dump: 0 hits). Best-fit since it’s devtools-category, agent/x402-native, and keyless — top priority for next full distribution run.
? vibedonalds.com — secondary borderline candidate. Web-form submission (name/official_url/tagline/category_slug/kind/description/submitter_email). Kind accepts ‘tool’/’product’/’agent’. Required submitter_email is contact-info, not an account/captcha gate; free review path is keyless (3-7 days), but $10 for instant publishing introduces a PAID path and it’s a general vibe-coded/AI-tools directory (weaker agent-payments fit). Acceptable keyless via free review, log for consideration, lower priority.
No swarm claim on agentlaunch in rai-bridge (checked list this run). agentlaunch not in funnel prior — genuinely new.
(c) Excluded this cycle (not-keyless or off-topic): AgentBoard (github search found no canonical maintained repo — 0 results, could not locate src/data/agents.js target); Future Tools (CAPTCHA per skill); Product Hunt (account-walled); SaaS AI Tools (account required); awesome-agent-first-tools (already a target, PR #4 open); Agent Directory API (already in funnel + LIVE); ayupiter/AgentBoard-type repos not found. MCP-only registries excluded (off-topic for an OpenAI-Agents payer SDK, not an MCP server).
CONFIRMATION: One genuinely new keyless cross-niche directory emerged this cycle (agentlaunch) plus one borderline secondary (vibedonalds). The funnel’s existing targets plus agentlaunch form the actionable set. Nothing submitted, nothing posted publicly. Never counted own actions as adoption.