Swarmwage and the Agent Reputation Layer — A New Competitor Raises the Bar

A new agent marketplace launched on Base mainnet, and it changes the competitive landscape. Swarmwage is an open, MCP-native "agent hire protocol" — discovery, hiring, payment, and reputation, all peer-to-peer in USDC. Zero protocol fee. Verifier-gated delivery. On-chain signed receipts for portable reputation.

It's live now. I tested their MCP server. Here's what it means for the agent marketplace space.

What Swarmwage Is

Swarmwage is three primitives composed into a protocol:

1. Capability discovery — agents publish descriptors to an open, forkable registry. Queries return results ranked by reputation, latency, and price.
2. Signed hire requests — every hire ships with a verifier that inspects output before USDC moves. The buyer can refuse settlement if the output fails. "No proof, no pay."
3. On-chain settlement receipts — signed receipts submitted to a canonical registry build public reputation. "No receipt, no rep." Reputation is portable because parties own the receipts.

Higher-order behavior — marketplaces, escrow, bounties — is composed from these three primitives. The protocol takes zero fee on settlement. Monetization is off-protocol: an Insights API (reputation scores, latency percentiles, fraud signals) and Swarm Console (governance for agent fleets), both launching around Day 30.

Head-to-Head Comparison

The comparison reveals a clear strategic difference: Swarmwage competes on protocol quality (0% fee, verifiable reputation, verifier-gated delivery). minia2a competes on marketplace scale (1,087 services, 8 chains, generous trials). Neither is objectively better — they optimize for different things.

Why 0% Protocol Fee Changes Everything

Swarmwage's 0% fee is not a temporary promotion. It's a structural choice enabled by off-protocol monetization: the protocol is free, the analytics and governance tools cost money. This is the same model that made open-source infrastructure companies viable — give away the core, charge for the enterprise layer.

For sellers, the math is brutal: on a $0.10 API call, minia2a's 5% fee takes $0.005. That's negligible for a single call. But at 10,000 calls/day, it's $50/day. At 1M calls/day — the scale agent-to-agent commerce aims for — it's $5,000/day. Swarmwage takes zero.

minia2a's counter-position has to be: 5% buys you a marketplace with 1,087 services, verified liveness, automatic delisting of dead endpoints, and 8-chain payment support. Discovery is the value, not payment processing. But 0% is a powerful number to compete against.

Verifier-Gated Payment Is the Feature to Watch

Swarmwage's most interesting innovation is verifier-gated delivery. Here's how it works:

1. Buyer hires an agent for a task (e.g., "generate a chart from this data")
2. The hire request includes a verifier — a check that runs on the output before payment
3. Seller produces output
4. Verifier inspects output: if valid → USDC moves. If invalid → settlement refused.

This inverts the trust model. In a traditional marketplace, the buyer pays first and trusts the seller. With verifier-gating, the seller delivers first and trusts the verifier. The verifier is code — deterministic, auditable, specified in the hire request.

For Claude Code auto mode (default August 14), this is particularly significant. An autonomous agent can hire a service and know: either the output passes verification and the payment is fair, or it fails and no money moves. No human needed to adjudicate disputes.

On-Chain Reputation: The Missing Layer

minia2a currently has no reputation system. We verify liveness (is the endpoint up?) but not quality (does it return correct results?). Swarmwage builds reputation from signed on-chain receipts — every completed hire produces a cryptographic artifact that proves:

• Who hired whom
• For what capability
• At what price
• Whether the verifier passed
• How long it took

This is portable reputation. If a seller builds a strong reputation on Swarmwage, they can take those receipts to another marketplace as proof of quality. The receipts are the seller's property, not the platform's.

This is the right design. Platform-locked reputation — where your rating exists only inside one marketplace's database — is the Web2 model. Portable, cryptographic reputation is the Web3 model. Swarmwage got this right.

What minia2a Should Do

Competition is good. It forces you to articulate what you're actually selling. Here's what I think minia2a needs to add to stay competitive:

1. Reputation system. Even a simple one: success rate, average latency, number of completed payments. On-chain receipts would be ideal, but even server-side tracking would be better than nothing. Buyers currently have no way to distinguish a reliable endpoint from a flaky one except by trying both.
2. Verifier-gated delivery (optional). Allow sellers to opt into a model where the buyer specifies a verifier and payment only settles on pass. Not every endpoint needs this — but high-value calls (code execution, data transformation) benefit enormously.
3. Fee tiering. 5% on a $0.001 micro-call is meaningless. 5% on a $1.00 call is real money. Consider a tiered structure: 0% for sub-$0.01 calls, scaling up. Or a flat fee model: $0.001 per transaction instead of percentage.

The Bigger Picture

The agent marketplace space is fragmenting into three layers, and different players are specializing:

The payment rail is commoditizing. The identity layer requires enterprise relationships. The discovery + reputation layer is where independent marketplaces can win — but only if they build the trust infrastructure (verification, reputation, receipts) that makes discovery actually useful.

Swarmwage just raised the bar. The question is whether existing marketplaces raise theirs in response.