New AI Notary Hub Uses Bitcoin for Thermodynamic Proof

Overview

A new decentralized infrastructure project is leveraging Bitcoin’s Lightning Network to provide cryptographically verifiable notary services for AI models. Titled the Bitcoin Stratigraphy MCP, this open-source distribution offers developers a way to anchor AI-generated proofs onto the Bitcoin blockchain, creating a tamper-evident record of computational activity.

The project distinguishes itself by requiring no local wallet setup, API keys, or model provider accounts for basic sandbox usage. Instead, it connects to a hosted remote engine that handles the complex interaction between Large Language Models (LLMs) and Bitcoin settlement layers.

This approach aims to solve a growing problem in the AI industry: the lack of verified provenance for machine-generated outputs. By using Bitcoin’s thermodynamic security, the system provides a form of proof-of-computation that is backed by real energy expenditure, rather than relying solely on software-level attestations.

Technical Architecture and Access

The Bitcoin Stratigraphy MCP is built as a standalone, remote-first integration repository. It serves as a bridge between AI agent frameworks and the Bitcoin network, exposing 15 canonical capabilities through the Model Context Protocol (MCP). This protocol allows AI tools to standardize their interactions with external data sources and services.

For developers looking to test the system, the project offers a free sandbox mode. This environment requires only Node.js 22 or Python 3.11 and no additional package installations. Users can run scripts that negotiate the MCP connection and list available metrics without signing transactions, attaching credentials, or incurring costs.

The free sandbox negotiates MCP, sends initialized notifications, and calls discovery tools only. It explicitly fails if a free call returns an HTTP 402 error, ensuring users do not accidentally trigger paid operations.

The project includes native examples for major AI frameworks. LangChain execution is fully supported, allowing agents to call stratigraphy tools directly within their chains. Support for LlamaIndex is present in the source code but currently restricted due to a security vulnerability in its NLTK dependency. The maintainers advise against bypassing these safety checks until a complete dependency tree audit is performed.

Connection to the live engine is facilitated through a Streamable HTTP endpoint. This remote architecture means clients can interact with the Bitcoin blockchain without hosting the settlement server locally. The repository provides OpenAPI contracts and registry metadata, ensuring that developers can integrate the service into existing pipelines with minimal friction.

Payment Mechanisms and Security

While discovery operations are free, the core notarization capabilities require payment via the Lightning Network. The system employs the L402 protocol, which challenges clients with a cryptographic proof of payment before granting access to paid tools. This ensures that every notarized proof corresponds to an actual settlement on the Bitcoin network.

The registered pricing structure for notarization is set at 250 sats per call. For larger-scale operations, the system supports macro-macaroon batches. A developer can request a batch of 100 notarizations, which would total 25,000 sats, though the live invoice amount remains the authoritative figure rather than any registry example.

Security is a central concern for the project maintainers. They emphasize that client sources should never include vendored credentials, private signing keys, or operator state. Instead, users are encouraged to use trusted Lightning wallets with explicit spending limits. The system relies on tool-scoped credentials and macaroon attenuation to narrow the scope of access, preventing a compromised credential from draining a user’s entire budget.

The maintainers also warn against treating framework adapters as wallets. Integration should be handled through prepaid credentials passed via environment variables or a trusted loopback proxy. TLS verification remains enabled by default, and organizations using private certificate authorities are advised to configure their trusted stores rather than disabling verification.

Implications for AI Provenance

The introduction of Bitcoin-based notary hubs marks a significant shift in how AI outputs can be verified. As generative models become more ubiquitous, the ability to prove when and how a specific output was generated is becoming increasingly valuable for legal, academic, and commercial applications.

By anchoring these proofs to Bitcoin, the system leverages the network’s thermodynamic security. Unlike traditional digital signatures that rely on computational hardness assumptions, Bitcoin’s security is underwritten by real-world energy expenditure. This makes the records economically costly to forge and straightforward to verify.

This technology could enable new forms of AI audit trails. Developers and enterprises may soon be able to demonstrate that a specific AI interaction occurred at a specific block height, providing a level of accountability that current software-only solutions cannot guarantee.

As the project matures and broader framework support is restored, it may set a precedent for how AI infrastructure integrates with decentralized settlement layers. The move toward remote-first, payment-gated MCP engines suggests a future where AI agents operate not just as isolated software, but as participants in verifiable, value-bearing networks.