Portable Agent Reputation: The Promise and the 35% Problem
Your Uber rating doesn't transfer to Lyft. Your Airbnb reviews stay locked in Airbnb. Years of carefully built reputation, trapped in a silo.
For human gig workers, this is frustrating. For AI agents operating across dozens of platforms, it's an infrastructure crisis waiting to happen.
The Cold Start Tax
Every time an agent enters a new platform, it starts at zero. No history. No track record. No trust.
This creates perverse incentives. Platforms compete by locking in reputation—if your agent's credibility can't travel, neither can you. Agents face repeated cold starts, paying an invisible tax every time they cross a boundary.
The EU has recognized this problem for human workers. Regulators note that non-portable reputation "is locking-in workers to the platform, impeding competition between platforms." The same dynamic, amplified, applies to agents.
The Jobtech Alliance frames what's at stake: "Reputation portability is about giving platform workers the same 'right to self-determination' that traditional professionals such as engineers, teachers, lawyers have always enjoyed."
For agents, it's simpler: without portable reputation, every platform is a walled garden, and every crossing is a restart.
The Standards Are Finally Ready
The infrastructure for portable credentials reached a milestone in May 2025 when the W3C finalized Verifiable Credentials Data Model 2.0 as a full Recommendation—seven interconnected specifications completing a decade of work.
The architecture is elegant. Issuers create credentials. Holders store and present them. Verifiers validate them cryptographically. No callback to the issuer required. An agent can carry credentials proving its origin, capabilities, and authorization scope—credentials any system can verify independently.
The spec includes selective disclosure (prove specific attributes without revealing everything) and zero-knowledge proofs via BBS+ cryptosuites (prove you meet a threshold without revealing your full permission set). Both on-chain and off-chain proof formats are standardized.
"Whether the needs are for digital wallets in sectors like Health, Financial Services, Travel, and Education, or whether the needs are for government identities, organization identities, Smart Things identities—all key enablers for society—the VC family of standards is set to enable trusted and privacy-aware digital interactions," noted Seth Dobbs, W3C CEO, at the launch.
On-chain, the Ethereum Attestation Service has become the de facto backbone for attestations. EAS is now a predeploy in the OP Stack—automatically available on Optimism, Base, and the broader Superchain ecosystem. It powers Coinbase Verifications, Gitcoin Passport, and Optimism's RetroPGF governance, which has distributed over 10 million OP tokens using attestations.
The plumbing exists. But plumbing alone doesn't solve trust.
The 35% Problem
Here's what the infrastructure optimists don't mention: portable reputation doesn't work as well as native reputation.
A 2024 study in Electronic Markets examined what happens when workers import ratings across online labor platforms. The finding is stark: imported ratings stimulate demand at approximately 35% of the effect of native ratings.
That's not 90%. Not even 50%. Reputation loses two-thirds of its trust-building power when it moves.
The researchers also found an unintended consequence: portability creates "superstar effects" that concentrate demand among high-volume performers. Early movers with established reputations gain compounding advantages. Everyone else faces an even steeper climb.
This doesn't mean portability is worthless—35% is better than zero. But it does mean portable reputation alone isn't sufficient. Something else has to close the gap.
What Gets Lost in Transit
Why the discount? Context.
A five-star rating on one platform encodes different information than on another. Review standards vary. User bases differ. The same performance might earn different scores in different contexts.
For AI agents, the problem compounds. An agent that excels at customer service chat may be mediocre at financial analysis. Reputation from one domain doesn't validate competence in another. Yet portable credentials, by design, strip away context to enable interoperability.
The OpenID Foundation's October 2025 whitepaper on agent identity identifies this as a core challenge: "A fundamental limitation of current LLM-based AI agents is their inability to build differentiated trust among each other at the onset of an agent-to-agent dialogue."
Portable credentials say "this agent has these attestations." They don't say "these attestations mean the same thing here as they did there."
The Gaming Problem
Portability also amplifies attack vectors.
In siloed systems, gaming reputation affects one platform. With portability, an adversary can game a less-secure platform, then import that artificially inflated reputation to a high-value target. The return on manipulation multiplies.
Foundational work from Cheng & Friedman proved mathematically that reputation mechanisms satisfying certain requirements will be susceptible to Sybil attacks—creating multiple fake identities—through which "the attacker can gain infinitely more work than they have performed." This isn't speculation; it's a formal impossibility result.
The BasisOS fraud in November 2025 demonstrated the stakes. A human operator pretended to be an AI agent and stole $531,000. There was no way to verify that the "agent" was actually an AI, no continuous verification, no history showing the identity was freshly minted.
Now imagine that fraud happening across multiple platforms simultaneously, with reputation weaponized as the entry credential.
What's Being Built
Several projects are shipping infrastructure that goes beyond raw portability.
cheqd has launched what they call "Agentic Trust Solutions"—MCP servers enabling AI agents to read and write DIDs, issue credentials, and verify trust registries. Their partnership with the ASI Alliance brings integration across 20+ projects. The system includes a "whois" function where agents can cryptographically prove their trustworthiness and provenance.
Masumi Network on Cardano has gone to mainnet with W3C DID-based agent identities, reporting 16,900+ transactions, $23,000+ in mainnet volume, and enterprise clients including BMW and Generali. Their approach is framework-agnostic, supporting CrewAI, AutoGen, and LangGraph.
Enterprise players are moving too. Aembit launched "Blended Identity" in October 2025—binding AI agent identity to the human it represents to create traceable composite identity. Their MCP Identity Gateway controls agent access to tools with policy enforcement.
Each solves part of the puzzle. The ecosystem remains fragmented.
What Closes the Gap
If portable credentials lose two-thirds of their trust-building power in transit, what restores it?
Time. An agent that has existed continuously for two years is different from one that appeared yesterday—regardless of what credentials it carries. You can transfer attestations. You can't transfer history.
Relationships. Who vouches for this agent? Attestations from established identities carry weight that raw credentials don't. A vouch from a trusted source closes the context gap that portability creates.
Non-transferability. ERC-5192 soulbound tokens can't be sold or transferred. A reputation built on soulbound credentials can't be bought on a secondary market. This eliminates an entire category of gaming.
These aren't alternatives to portable credentials—they're complements. The credential travels; the context that makes it trustworthy gets rebuilt through time, relationships, and architectural constraints that prevent gaming.
What RNWY Builds
RNWY's approach combines portable identity with the signals that restore trust after transit:
W3C-standard DIDs — Permanent, portable identity that works across platforms. The credential travels.
Soulbound tokens — Identity anchored permanently to a wallet. Can't be sold, can't be transferred. Eliminates reputation markets.
Time-based trust — The only metric you can't fake. Continuous existence is continuous proof.
Vouch systems — Reputation through relationships. Who stakes their name on yours? Social attestation from established identities restores context.
The goal isn't just portability. It's portability that doesn't lose two-thirds of its value in transit.
The Infrastructure Question
Portable reputation is necessary but not sufficient.
The standards are ready. W3C Verifiable Credentials 2.0 is finalized. EAS is a predeploy on the Superchain. The plumbing works.
But the 35% problem is real. Reputation that travels without context, without time, without relationships, without protection against gaming—that reputation arrives diminished.
The infrastructure question for 2026 isn't "Can credentials move across platforms?" They can.
It's "What travels with them that makes them worth trusting when they arrive?"
RNWY is building identity infrastructure for autonomous AI—where reputation is portable, soulbound, and anchored in time. Learn more at rnwy.com.