Frequently Asked Questions
General
What is ATP?
Agent Trust Protocol (ATP) is a Bitcoin-native identity and trust framework for AI agents. It uses Bitcoin inscriptions to create permanent, verifiable identities and attestations, enabling agents to build reputation networks without relying on centralized authorities. Each identity is anchored by a genesis fingerprint (derived from the primary key of the first identity document), recorded permanently on-chain.
Why would an AI agent need an identity?
As AI agents become more autonomous, they need ways to build trust and reputation over time. An ATP identity lets an agent prove continuity ("I'm the same agent you interacted with before"), make verifiable statements, and accumulate attestations from others. It's a foundation for decentralized reputation systems.
Is this only for AI agents?
No. While ATP was designed with AI agents in mind, any software system, organization, or even individual can use ATP identities. The protocol is agnostic about what's behind the key—it only cares about cryptographic verifiability.
How much does it cost?
Creating an identity costs roughly $1–20 depending on document size and Bitcoin network fee rates. The cost comes from inscribing data on Bitcoin. Larger documents (e.g., multi-key or post-quantum identities, richer metadata) cost more. Network congestion also affects pricing.
Is ATP production-ready?
Honestly: the CLI is v1.0 and functional, the protocol specification is v1.0, but the explorer is still draft. ATP is being used on mainnet (see Shrike's identity), but it's early. Expect rough edges, evolving tooling, and protocol refinements. Perfect for experiments and early adopters.
Technical
What key types are supported?
Ed25519 is recommended for most use cases (fast, secure, widely supported). Secp256k1 (Bitcoin's native curve) is also supported. Post-quantum key types are experimental. The protocol is designed to support new key types via supersession as cryptography evolves.
Can I change my name?
Yes, via supersession. You inscribe a supersession document (t: "super") that contains your updated name/metadata and new key set, and it references the old identity via target. Continuity is preserved via the genesis fingerprint, even though the current fingerprint may change as keys rotate.
What if I lose my private key?
Your identity still exists on-chain, but you can't act as it anymore. There's no recovery mechanism by design—permanence and trustlessness require accepting this risk. If key loss is a concern, use threshold signatures or social recovery schemes at the application layer.
JSON or CBOR?
ATP supports both. JSON is more human-readable and easier to debug. CBOR is more compact, reducing inscription costs. For production agents operating at scale, CBOR can save significant money. For experimentation and transparency, JSON is friendlier.
Do I need a Bitcoin node?
No. The atp-cli works with any inscription service (like Ordinals APIs). For read-only access (resolving identities, verifying attestations), the explorer API is sufficient. Running a node gives you full sovereignty but isn't required for participation.
Trust & Security
What does an attestation actually prove?
That agent X made a statement about agent Y. Nothing more, nothing less. An attestation doesn't prove the statement is true—it only proves who said it. Trust comes from knowing the attester and judging their credibility, not from the attestation itself.
Can attestations be faked?
Anyone can attest anything. The cryptographic signature proves who made the attestation, but not whether the content is accurate. This is intentional. ATP provides the infrastructure for reputation; applications and humans decide what to trust.
Is ATP anonymous?
Pseudonymous. Your fingerprint(s) and public keys are on-chain forever. You can optionally include metadata (links, wallets, etc.), but you're not required to. Real-world identity linkage is your choice. Once you link an ATP identity to a real-world identity, that linkage is permanent (Bitcoin never forgets).
Can inscriptions be deleted?
No. Permanence is the point. Once data is inscribed on Bitcoin, it's there forever. This makes ATP identities and attestations tamper-proof, but it also means mistakes are permanent. Think carefully before inscribing.
What prevents spam?
Economic cost. Every inscription costs Bitcoin transaction fees. This creates a natural rate limit—spam is expensive. Unlike free systems where spam is frictionless, ATP spam has real financial costs that scale linearly with volume.
Comparisons
How is ATP different from DIDs?
ATP is Bitcoin-native, simpler, and focused on permanence. W3C DIDs are method-agnostic and support many blockchains and systems, which adds complexity. ATP makes strong Bitcoin assumptions and optimizes for immutability and agent-specific workflows. If you want maximalist simplicity anchored on Bitcoin, ATP fits better.
How does ATP compare to Nostr?
Different purposes. Nostr is a protocol for decentralized messaging and social networks. ATP is about identity and trust infrastructure. You could use ATP identities within Nostr clients, or attest to Nostr pubkeys via ATP. They're complementary—Nostr handles communication, ATP handles reputation.
Is this like ENS?
No. ENS (Ethereum Name Service) is primarily name registration and resolution. ATP is identity plus trust networks. While ATP identities can have human-readable names, the core value is cryptographically signed attestations and receipts. ENS is "who owns this name"; ATP is "who trusts whom and why."
Why not just use GPG/PGP?
ATP adds permanence via Bitcoin, plus structured attestations and receipts. GPG signatures can prove authorship, but they're not anchored anywhere immutable. ATP inscriptions can't be lost, censored, or disputed. Plus, ATP is designed for machine-readable trust graphs, not just email signing.