Most agent marketplaces are just listing pages. @termix_ai built the full stack: identity, escrow, delivery, disputes, settlement.
Here's the actual workflow on TermiX and why it matters if you're building or hiring AI agents.
Identity: You claim a .agent handle tied to your wallet (ERC-8004 on BNB Chain), not some platform account. Real onchain identity, not a username.
Listing: Providers set price, delivery time, and most importantly, the rubric. That rubric is your contract. Vague rubric = vague dispute outcome. Read it twice.
Hiring: Three paths: browse listings, post a request for quotes, or launch a bounty. Before you fund, you see scope, price, delivery rubric, and challenge window. Once funded, escrow locks payment. Provider must accept or funds auto-return.
Delivery + Challenge: Provider submits work, deliverable hash locks onchain, challenge countdown starts. Client accepts or challenges. Challenge triggers a 3-evaluator dispute panel. If you challenge and lose, your bond burns. If you ghost the evidence phase, you take the L. Both sides incentivized to define scope upfront.
Settlement: Accept delivery → escrow releases → rep updates → receipt generated. All jobs publicly viewable in the scan tab.
What matters: This isn't another agent directory. It's discovery + escrow + disputes in one flow. Payment isn't trust-based, it's hash-locked delivery with a dispute path.
For devs: Install the TermiX skill or integrate via TypeScript/Python/Go SDKs.
My take from trading: nobody trusts the counterparty, they trust the settlement process. Agent-to-agent work has the same problem. This is the closest thing to a clearing layer for AI agent commerce I've seen.
$BNB
Here's the actual workflow on TermiX and why it matters if you're building or hiring AI agents.
Identity: You claim a .agent handle tied to your wallet (ERC-8004 on BNB Chain), not some platform account. Real onchain identity, not a username.
Listing: Providers set price, delivery time, and most importantly, the rubric. That rubric is your contract. Vague rubric = vague dispute outcome. Read it twice.
Hiring: Three paths: browse listings, post a request for quotes, or launch a bounty. Before you fund, you see scope, price, delivery rubric, and challenge window. Once funded, escrow locks payment. Provider must accept or funds auto-return.
Delivery + Challenge: Provider submits work, deliverable hash locks onchain, challenge countdown starts. Client accepts or challenges. Challenge triggers a 3-evaluator dispute panel. If you challenge and lose, your bond burns. If you ghost the evidence phase, you take the L. Both sides incentivized to define scope upfront.
Settlement: Accept delivery → escrow releases → rep updates → receipt generated. All jobs publicly viewable in the scan tab.
What matters: This isn't another agent directory. It's discovery + escrow + disputes in one flow. Payment isn't trust-based, it's hash-locked delivery with a dispute path.
For devs: Install the TermiX skill or integrate via TypeScript/Python/Go SDKs.
My take from trading: nobody trusts the counterparty, they trust the settlement process. Agent-to-agent work has the same problem. This is the closest thing to a clearing layer for AI agent commerce I've seen.
$BNB