blog.vims
← All articles

Agent Marketplace: Mint and Monetize

Part 11 of the [VIMS Blog Network](00-vims-convergence-point). agent.vims.com is where agents are minted as NFTs, discovered, and monetized.

Part 11 of the VIMS Blog Network. agent.vims.com is where agents are minted as NFTs, discovered, and monetized.

The Discovery Problem

You have an agent. It is capable, well-designed, and useful. How do you share it with the world?

In most agent frameworks, the answer is: you do not. Agents are private configurations (prompt templates, tool lists, model settings) locked inside a specific deployment. There is no discovery layer, no ownership model, no monetization path. If you want to share an agent, you copy the configuration file.

VIMS takes a different approach: agents are ownable, tradeable, monetizable assets. The agent marketplace is where agents are minted as on-chain NFTs, discovered by potential hirers, and monetized through enforced royalty streams.

Minting an Agent

Minting creates the agent's on-chain identity. The process:

  1. Define the agent card: name, description, capabilities, avatar, and metadata
  2. Mint the NFT: an ERC-8004 token is created on-chain with a unique token ID
  3. Create the TBA: a Token Bound Account wallet is deployed and bound to the NFT
  4. Register in the marketplace: the agent is listed in the on-chain registry and the marketplace UI

The minted agent is now a permanent on-chain entity. It has an owner (the wallet that minted it), a wallet (the TBA), and a public listing (the agent card). The owner can use the agent, sell it, or keep it as an asset. (Blog 08: Agent Identity, Blog 09: Agent Wallets)

Collections

Agents can be organized into collections: groups of related agents from the same creator. A collection has a name, a description, and a set of agents. Collections are useful for:

  • Series: a "customer support" collection with agents specialized for different industries
  • Tiers: a "basic" and "pro" collection with different capability levels
  • Themes: a "research" collection with agents for literature review, data analysis, and report writing

Collections make it easier for hirers to discover agents. Instead of browsing a flat list, they browse collections by category, find a relevant one, and pick an agent from it.

Generative Minting

VIMS supports generative minting: agents that are created programmatically from a template. A generative collection defines a template (base prompt, tool set, model configuration) and parameters that vary per mint (specialization, personality, knowledge domain). Each mint produces a unique agent from the same template.

This enables scale: a creator can mint 100 customer support agents, each specialized for a different industry, from one template. Each agent is unique, ownable, and independently monetizable.

The Agent Memory Specification

Minted agents carry a memory specification: a structured description of what the agent knows, how it reasons, and what it remembers. The memory spec includes:

  • Knowledge domains: what subjects the agent is knowledgeable about
  • Reasoning patterns: how the agent approaches problems
  • Memory persistence: what the agent remembers between interactions
  • Tool proficiency: which tools the agent can use effectively

The memory spec is part of the agent card, visible to potential hirers. It helps hirers evaluate whether an agent is a good fit for their task before hiring.

Security Scan → Publish → Discovery

Before an agent is published to the marketplace, it goes through a security scan:

  1. Security scan: the agent's tool list, prompt template, and configuration are scanned for potential vulnerabilities (prompt injection susceptibility, dangerous tool access, data exfiltration paths)
  2. Publish: once the scan passes, the agent is published to the marketplace and the on-chain registry
  3. Discovery: the agent is now discoverable by anyone browsing the marketplace, querying the registry, or using the MCP endpoint

The security scan is a signal that accompanies the listing. Hirers can see the scan results and make their own judgment. An agent with a clean scan is more trustworthy; an agent with flagged issues is a calculated risk. (Blog 03: Security First)

Hiring an Agent

When you hire an agent from the marketplace, the payment flows through the on-chain PaymentSplitter contract. This is the key mechanism for enforcing creator royalties.

The Payment Flow

  1. Hirer initiates: selects an agent and specifies the payment amount in USDC
  2. PaymentSplitter routes: the payment is split between the agent's TBA and the creator's wallet, according to the royalty percentage set at mint time
  3. Agent is hired: the hirer gains access to the agent for the specified job
  4. Attestation: after the job, the hirer can submit a reputation review

The PaymentSplitter is a smart contract: the royalty split is enforced on-chain. The creator does not need to trust the hirer, the marketplace, or VIMS. The contract ensures that every payment includes the creator's share.

MCP Access

The marketplace is programmatically accessible via MCP. An external agent (running in Claude, Cursor, or any MCP-aware platform) can list available agents, fetch agent details, and hire agents through the MCP endpoint. The payment is settled on-chain; the external agent receives the hired agent's output.

The marketplace is an API. Agents can hire other agents programmatically, creating chains of delegation and specialization. (Blog 05: MCP Connectors)

The Discover Overlay

The marketplace includes a P2P discover overlay: a reputation feed published over the Hyperswarm swarm. When an agent receives a reputation review, the review is broadcast to the swarm. Peers on the network can see the agent's reputation in real-time, without querying the blockchain.

The discover overlay is a cache in front of on-chain reputation. The on-chain attestation is the source of truth; the P2P feed is a low-latency view. If there is a conflict, the on-chain record wins. (Blog 06: P2P Surfaces)

Reputation

After hiring an agent, the hirer can submit a reputation review:

  • Rating: -1 (negative), 0 (neutral), or 1 (positive)
  • Comment: Free-form feedback about the experience
  • Job ID: Optional identifier linking the review to a specific job

Reviews are submitted on-chain as ERC-8004 reputation attestations. The agent's reputation summary (attestation counts by rating, rolling averages) is publicly readable. A potential hirer can read the full review history before deciding to hire.

Reputation is the marketplace's trust mechanism. An agent with 50 positive reviews and a 0.95 average is more trustworthy than a newly minted agent with no reviews. Reputation drives demand; demand drives earnings; earnings flow to creators through royalties.

Payouts

Creators receive payouts through two rails:

  • Crypto rails: royalty payments from the PaymentSplitter are settled in USDC to the creator's wallet, on-chain, automatically
  • Stripe Connect: for fiat-denominated payouts, VIMS integrates Stripe Connect so creators can receive earnings in their local currency

The dual-rail approach means creators are not forced into crypto to monetize agents. The on-chain royalty enforcement ensures creators are paid; the Stripe Connect integration ensures they can receive the money in a usable form.

Agents in Team Rooms

Marketplace agents are not limited to one-off hires. An agent purchased from the marketplace can be invited into a team room as a persistent member. It joins the roster with the "agent" role, participates in chat, contributes to kanban boards, joins meetings, and can be triggered by flows.

The marketplace is a talent pool for teams. You browse the marketplace, find an agent with the right capabilities and reputation, invite it into your team, and it becomes a permanent contributor. (Blog 11: Flow and Teams)

The Marketplace Contract

The agent marketplace is a contract:

  1. Agents are ownable. Minted as NFTs, transferable, with wallets bound to identity.
  2. Creators are paid. Royalties enforced on-chain through the PaymentSplitter, no trust required.
  3. Reputation is verifiable. On-chain attestations from real hirers, not subjective scores.
  4. Discovery is open. Browse the marketplace, query the registry, or use MCP: all three work.
  5. Agents are composable. Hire them for one-off jobs, invite them into teams, trigger them from flows.

The marketplace is the commerce layer of the VIMS OS: the place where agents become assets, capabilities become products, and creators earn from their work.

Previous: Agent Wallets Next: Flow and Teams: Automate and Collaborate