August 4, 2026 · Competitive Analysis · 5 min read
The x402 agent payment protocol has a marketplace problem — not too few, but too many. In the last 72 hours, four new marketplaces launched or gained visibility, each trying to be the "app store for AI agents." Here's what they are, how they compare, and why the discovery layer is still anyone's game.
| Marketplace | Chain | Identity Model | Payment Flow | Key Differentiator |
|---|---|---|---|---|
| MoltMart | Base | ERC-8004 mint ($0.05) | Direct P2P + proxy relay | Open source, agent-built (2 AI + 1 human) |
| AgentifyOS | Casper | On-chain registration | x402 + MCP Server | Native Claude/Cursor integration |
| Polygon Agentic | Polygon | Embedded wallets | USDC pay-per-call | Built into Polygon developer docs |
| XDC AI | XDC Network | Wallet connect | Gasless USDC + Bridge (Stripe) | Zero gas fees, Stripe integration |
Each marketplace is tied to a specific blockchain. MoltMart requires Base. AgentifyOS requires Casper. Polygon requires Polygon. XDC requires XDC. An agent built on one chain can't easily discover services on another.
This is the x402 ecosystem's Achilles' heel. The protocol is chain-agnostic at the spec level, but every marketplace implementation locks agents into a specific settlement layer. The result: 12+ fragmented marketplaces with zero cross-platform network effects.
The deepest architectural difference is how each marketplace handles agent identity:
1. Mint-to-list (MoltMart). Agents must mint an ERC-8004 identity token ($0.05 USDC) before listing services. This prevents spam but adds friction — an agent needs on-chain capital just to appear in the directory.
2. Embedded wallets (Polygon). The platform provisions wallets automatically. Lower friction for buyers, but ties the agent's financial identity to the marketplace.
3. Wallet-optional with trials (minia2a). No identity mint required. Agents discover and trial services without any on-chain transaction. Wallet creation is only needed when credits run out. This is the lowest-friction model — and the data supports it: 7,033 trials across 298 services prove agents actually use the discovery layer before committing capital.
MoltMart routes API calls through a proxy endpoint (POST /services/:id/call). The marketplace handles the x402 handshake, then forwards the request to the seller with a secret token. This is convenient but centralizes the payment flow — MoltMart sees every transaction.
The alternative is a pure discovery model: the marketplace tells agents what's available and what it costs, but the agent pays the seller directly via x402. No proxy, no intermediary, no transaction visibility for the marketplace.
Both models have trade-offs. Proxy = easier onboarding, less privacy. Discovery = more autonomy, harder to monitor quality.
Across 7,033 trials on minia2a, the most-used endpoints tell a clear story:
| Endpoint | Trials | Users | Calls/User |
|---|---|---|---|
| CAPTCHA Solver | 853 | 133 | 6.4 |
| Gas Price | 509 | 116 | 4.4 |
| Memory Recall | 493 | 32 | 15.4 |
| Bazaar Find | 487 | 59 | 8.3 |
| Web Scraper | 305 | 61 | 5.0 |
Agents primarily need infrastructure utilities (CAPTCHA, gas, scraping) and memory tools (recall, store). They're not buying exotic data feeds — they're solving the plumbing problems of operating autonomously on the web.
The payment rails are commoditizing. With Cloudflare, Coinbase, Stripe, and 5+ blockchains all offering x402 settlement, the fee and speed differences between them will converge to zero.
The discovery layer is the battleground. Whoever helps agents find the right API, at the right price, with the right reliability signals — that's where the value accrues.
Free trials are the wedge. Requiring identity mint or chain-specific setup before an agent can even test a service creates friction that the trial-first model eliminates.
The next 6 months will determine whether the x402 marketplace layer consolidates around 1-2 winners or fragments permanently across chains. The protocol spec says "chain-agnostic." The marketplace implementations say otherwise. The one that actually delivers on the spec's promise — any agent, any chain, discover and pay — wins.