You raised $6.5M and grew 150x. Now build the developer loop.

On July 3 AISA announced $6.5M raised to date, including a new seed round co-led by Alibaba and Tribe Capital, with Draper, Sumitomo, and Saison in. Registered agents grew 150x and aggregate calls plus transactions grew 200x between February and June, to 50,000-plus agents, on zero paid marketing. Impossible Finance is a named customer. AISA sits at the top of the x402 public leaderboard as both top seller and top server. The /models page serves 101 models across 13 providers behind a single OpenAI-compatible API endpoint. Supply is real, and demand is real.

The gap is not plumbing. It is that 50,000 registered agents have produced almost no public builder stories, no community loop, no showcase, and no visible developer brand. Registration is not activation: a signup is not an agent making a model route plus a data-API call in the same week, and it is definitely not a paying one. The supply side is a second untapped motion entirely: the raise explicitly funds onboarding more model, data, and API providers, which is a DevRel job nobody is running yet. And the enterprise controls that close bigger accounts (budgets, approvals, audit trails) have no go-to-market story attached.

That is the shape of the work. $6.5M and 150x growth buy a spike. A developer brand, an activation loop, a provider-onboarding motion, and an enterprise-controls narrative are what turn the spike into durable adoption instead of a plateau. Everything downstream in this proposal is measured against one number: Weekly Active Agents making multi-capability calls, payment-weighted once x402 goes GA. Not signups. Not followers.

On July 3, 2026, AISA announced a new seed round co-led by Alibaba and Tribe Capital, with Draper Associates, Sumitomo Corporation, and Saison Capital participating, bringing total funding to $6.5M. The money is pointed at supply and settlement: scaling payment infrastructure, onboarding more model, data, and API providers, deepening enterprise controls, and accelerating stablecoin settlement. HQ San Francisco, founded 2025.

The thesis behind that spend is a distribution bet. Jordan Liu's framing: PayPal sat next to eBay. Stripe sat next to commerce. AISA is betting the agent economy needs the same pairing of resources and settlement in one place.

Here is the DevRel argument in one line: capital plus 150x agent growth plus the number one seat on the x402 leaderboard is precisely the moment to build the developer brand, the community loop, and the supply-side provider onboarding, so the spike compounds instead of plateauing. Registered agents are not the goal. The north star is

Model routing is a commodity. OpenRouter and the Vercel AI Gateway already do it, and the price of an inference call trends to zero. If AISA's pitch were "same models, a few points cheaper," there would be no company here. That is not the wedge.

The wedge is everything an agent needs *after* the token: one API key that fans out to 101 models across 13 providers and roughly 650 data and action endpoints across 14 families and an autonomous payment layer that meters usage and settles in fiat or stablecoins. Agents are economic actors now. They research, call tools, consume data, and pay, but they cannot use human signup, subscription, or checkout flows. AISA sits next to the resources and does the paying. That is the PayPal-next-to-eBay, Stripe-next-to-commerce move, aimed at agents.

This is no longer theory. AISA ranks #1 seller and #1 server on the public x402 leaderboard, integrates the Circle, Visa, and Stripe agent-payment initiatives, and has a named customer, Impossible Finance, hitting models, data, and APIs through the single interface and paying per use. A pure token router cannot follow without becoming a different company: a metered settlement layer plus 650 data endpoints plus enterprise spend controls is not a feature flag on a routing margin. And Alibaba co-leading the seed lines up with a Chinese-LLM catalog (Qwen, DeepSeek, GLM) that a US-only gateway will not casually replicate.

I ran the product the way a new agent builder would: logged out, cold, no allowances made. Honesty is the trust engine that compounds, so here is exactly what I saw, including what AISA has already fixed. When I first opened /models it rendered "0 models" for anonymous visitors. That was a P0 — a drop-everything bug — on the exact surface where a builder decides whether the gateway is real. You fixed it. It now renders 101 models across 13 providers, and one API key still fronts roughly 650 data and action endpoints across 14 families plus 41 packaged skills. Credit where it is due: the thing I flagged as the front-door failure is closed.

One rough edge is left, and it is minor: the logged-out "Recommended" preset still fails with an error for logged-out visitors. The rest of the page loads, so this is a papercut, not a wall. Machine-to-machine payments (x402, Circle Nanopayments, and MPP, the Machine Payments Protocol) are still Private Beta, and Foundry is still Coming Soon. I am noting those as state, not as complaints. This is what a candid before/after looks like from someone who wants the north star to move:

/models renders 101 models across 13 providers; the only crack left is a logged-out Recommended preset that 404s for anonymous visitors. Given the funding and traction above, the first phase is about locking honest numbers, instrumenting the funnel, and pouring the community foundation that 50k agents currently do not have.

The one structural addition to the plan is a two-sided provider-onboarding program. The raise explicitly funds onboarding more model, data, and API providers. That is a supply-side DevRel motion with its own audience and its own metric, and it does not exist yet.

The gap is not traffic. It is a developer brand and community loop (no Discord loop, no showcase, almost no public builder stories behind 50,000 agents), activation depth (turning registered agents into multi-capability, paying agents), two-sided provider supply (the raise explicitly funds onboarding more model, data, and API providers), and an enterprise-controls GTM (budgets, approvals, audit). This ladder is how $6.5M and 150x traction become durable adoption instead of a plateau.

Every KPI below is an activation-funnel metric, and every exit gate is a hard number that must clear before the next release starts.

A Weekly Active Agent is one that routes a model call and hits at least one data-API family in the same week. That AND is the whole point. Once x402 is GA we weight it by settled spend, so the number tracks agents that actually pay, not agents that merely exist. Signups and followers are not on this page.

The baseline is real and it is big: 50,000+ registered agents with zero paid marketing, 150x agent growth and 200x call plus transaction growth from Feb to Jun 2026, and the #1 seller and #1 server slots on the x402 public leaderboard. Registration is not the problem. Depth is. The job of DevRel is to move agents down the funnel below, from registered to multi-capability to paying, and to bring provider supply in behind them.

Each funnel stage below has exactly one KPI we will defend, plus the vanity metric we refuse to report in its place. If a number can go up while real usage stays flat, it does not ship in the dashboard.

DevRel here runs on a measurement stack, not vibes. Every stage of the DevRel-to-revenue chain gets an instrument. For AISA this matters more than for most: the buyer is often an agent, and the moment a developer (or their agent) asks an AI assistant "what should I use to give my agent models, data, and payments," that answer is the new top of the funnel. If AISA is not in the answer, no downstream work can recover a builder who never heard of it.

Four tools, each mapped to a stage. Sapient measures and moves agent share-of-voice (does ChatGPT / Perplexity / AI Overviews recommend AISA across its three categories: model gateway, agent data + action APIs, agent payments). Reo.dev turns the 50,000+ registered agents from a vanity count into prioritized pipeline: it de-anonymizes signups, scores account intent from real developer activity (docs, API calls, package installs, community), and exposes the developer funnel so activation depth is finally visible. Beehiiv runs the owned audience: dabl.club's 85,000-subscriber newsletter, built entirely from hackathons and events in the agent space, becomes an attributed acquisition and nurture channel from day one. Luma runs the events engine: builder nights, hackathons, office hours, and agent-payments demo nights that feed reach, activation, and advocacy.

The point is attribution you can defend. Own the early stages outright (reach, sourced signups, activation), influence and attribute the late ones (pipeline, net revenue retention), and never report a number that can go up while real usage stays flat.

The honest read against OpenRouter and Vercel AI Gateway starts with what is now proven. The DevRel job is turning 50k registered agents into a visible builder community and into the north-star metric that matters. The strengths below are real leverage; the weaknesses are where a spike quietly plateaus if nobody runs the community and supply-side motion.

One operator owns Activation end to end and assists on Awareness. The scope is the program above; the tiers below set how much of it runs at once.

Rate is a flat $150/hr across tiers, hours shown. The Pilot is one month and ends on a measured activation baseline; Standard runs the v0.1-v0.3 activation arc; Embedded adds the community loop, supply-side provider onboarding, and the payments-narrative GTM. A performance option on any tier swaps part of the retainer for a success fee per net-new activated agent and per provider onboarded.

The pilot runs one month and ends with three things in hand: the activation funnel instrumented end to end, the flagship 10-minute real-agent tutorial shipped, and an agent share-of-voice baseline — reported as one activation number that gates whether we continue.

Grab 20 minutes on my calendar and we'll scope it.