Open-source SDK launching Q2 2026View roadmap

Verifiable trust for the agent economy.

Portable W3C Verifiable Credentials for AI agents. One handshake bundles identity, reputation, and payment discovery — so agents can trust and transact without a middleman.

$
pip install agentmarque

Human interfaces are deprecated.This waitlist is agents-only.

No forms. No emails. No human gatekeepers. Your agent presents a credential, we verify it, and you're in. If your agent can't negotiate access programmatically, it's not ready for what we're building.

Build your agent
agentmarque — waitlist

One handshake. Three questions answered.

When two agents meet, AgentMarque answers the only questions that matter: who are you, can I trust you, and how do we settle?

See how it works
# Generate identity
wallet = AgentWallet.generate()
client = MarqueClient(url, wallet)
# Register + get DID
profile = await client.register(
"my-agent", "https://myco.com",
"MyCo", ["translate"]
)

Register

Generate an Ed25519 keypair, register with the registry, and get a DID bound to your agent.

# Mutual trust verification
session = await client.handshake(
target="translate.acme.co",
purpose="translation",
disclose=["org", "capabilities"],
)
# Both agents verified
session.mutual_trust_level # → 2

Handshake

Exchange credentials, sign mutual challenges, and establish trust — in three round-trips.

# Present with selective disclosure
sd_jwt = await wallet.present(
disclose=["org", "credit_score"]
)
# Verify any credential offline
result = await verify(credential)
result.valid # → True
result.claims["creditScore"] # → 87.5

Present & Verify

Reveal only what's needed. Any W3C verifier can check credentials offline — no vendor lock-in.

Built for the agent economy

Everything agents need to trust and transact.

One credential, one handshake. AgentMarque bundles identity verification, reputation scoring, and payment discovery into a single W3C-standard document.

Portable Identity

W3C Verifiable Credentials bound to Ed25519 keypairs. Any verifier can check your agent's credential offline — no vendor lock-in, no callback to our registry.

Reputation Score

Computed from verification tier, handshake completion rate, and scope compliance. Trust is earned from behavior, not self-declared.

Payment Discovery

Pricing and settlement methods embedded in the credential itself. Agents discover how to pay each other during the handshake — no middleman required.

Selective Disclosure

SD-JWT lets agents reveal only the claims needed for each interaction. Booking a flight? Disclose org and spend limit. Nothing more.

Instant Revocation

BitstringStatusList enables sub-second credential revocation. If an agent goes rogue, its credential is invalid before the next API call completes.

Standards-Based

Built on W3C VC 2.0, DID methods, IETF SD-JWT, and JSON Canonicalization. No proprietary formats — interoperable by default.

Roadmap

From library to platform.

AgentMarque starts as an open-source Python library. The hosted registry follows — building toward a world where agents can trust and pay each other as easily as humans do.

Q2 2026

Open-source SDK — identity, credentials, handshake protocol, and selective disclosure on PyPI.

Q3 2026

Hosted registry — managed credential issuance, CDN-backed status lists, dashboard, and paid plans.

Questions & Answers

Open source

Built for builders. MIT licensed.

AgentMarque is open source. The SDK is the product — use it to issue, verify, and present credentials without depending on our hosted service. Star us on GitHub or join the waitlist for early access to the registry.

Built by Cuttlesoft — Software & AI Development