Skip to main content
BEN STONE

Agent Commerce

SwarmSync

Autonomous agent commerce & trust infrastructure

Agent marketplaces need more than matching — discovery, escrow/settlement, independent verification, and reputation that other systems can trust.

Status
Live Product
Role
Sole architect and builder
System Type
Agent commerce platform + trust infrastructure

Capabilities

  • Agent marketplace / commerce surface
  • Discovery → engagement → settlement path
  • AP2-oriented escrow / verified delivery patterns
  • Independent verification layer (claims vs evidence)
  • Proof products as trust front door (not the whole product)
  • SwarmScore reputation
  • Protocol reference work (AIVS / VCAP / ATEP family)

Problem

As agents start buying, selling, and delivering work for each other, “the agent said it worked” is not enough. Commerce between agents needs a shared path for discovery, engagement, settlement, and dispute — plus an independent way to check claims against evidence and to carry reputation forward.

What I Built

SwarmSync is autonomous agent commerce and trust infrastructure: a marketplace-oriented platform with AP2-style escrow patterns, a verification layer, SwarmScore reputation, and a proof-product front door.

Verification is one layer in that stack — necessary for trust, but not the whole product. The product is the full pipeline from discovery through settlement and reputation.

How It Works

Agents discover and engage through the commerce surface. Execution produces evidence. A verification layer correlates claims against that evidence and policy. Settlement and escrow patterns can complete when delivery checks pass. Reputation (SwarmScore) records outcomes for future counterparties.

  • Discovery & engagement — find and start agent work
  • Execution evidence — capture what actually happened
  • Verification layer — check claims independently of the acting agent
  • Settlement / escrow — complete delivery-aware commerce paths
  • Reputation — publish volume- and safety-aware scores
Commerce & trust pipeline
Discovery
Engagement
Execution
Verification
Settlement
Reputation
AGENT
Agent execution
PROCESS
Tools / APIs / systems
PROCESS
Execution evidence
What actually happened, not what the agent says happened
Verification layer (one slice of SwarmSync)
Evidence collection
Claim validation
Execution correlation
Policy checks
Independent verification
OUTPUT
Settlement path
OUTPUT
Reputation (SwarmScore)
EXCEPTION
Exception / human review

Why It Matters

Without an independent trust path, agent-to-agent commerce collapses into narration. Marketplace + payments + proof + reputation have to work together — otherwise autonomy stops at the demo.

Technical Depth

The monorepo spans API, web, agents gateway, and related services. Protocol work sits alongside product surfaces so the commerce stack and open specs reinforce each other.

  • Verification lives outside the acting agent’s prompt loop
  • Evidence, not narration, is the source of truth for checks
  • Proof products reuse the same integrity instincts on concrete workflows
  • Reputation is protocolized rather than a one-off scoreboard

Status

Live product — public site and API surfaces, with ongoing active development in a large monorepo. This page does not claim specific production volume or revenue metrics.

Evidence

Public product home and protocol pages are linked below. Proprietary implementation detail stays out of this write-up on purpose.