
Rubric Protocol
io.github.0xsimsPublished by GitHub account 0xsimsv2.0.2Updated Oct 24, 2026
AI compliance attestation for EU AI Act, SR 11-7, HIPAA. Free local tier, no key required.
Overview
Local MCP server that lets an assistant create, verify and export signed compliance attestations for AI decisions, with optional Hedera mainnet anchoring.
- What you can do
- Create and verify attestations for AI decisions, including batch attestation, pipeline traces and proof retrieval, with optional Merkle-chained anchoring to Hedera mainnet and zero-knowledge verification tools. Generate compliance evidence such as EU AI Act Annex IV documents, C2PA assertions, W3C verifiable credentials, compliance reports and regulatory filings. Register agents and models, map and assess jurisdictions, and certify against NIST AI RMF. Record governance activity such as incidents, human reviews and adversarial sessions, and produce usage reports and auditor tokens. Tools are grouped into modules; the default profile exposes eight core tools and more can be enabled through a
- When to use it
- Use it when you need a verifiable, exportable audit trail of AI decisions for frameworks such as the EU AI Act, SR 11-7, HIPAA or NIST AI RMF, or when you want to try local signed attestation without an account.
- Requirements
- Runs locally as an npm package started with npx, so Node.js is needed. The free local tier needs no key; anchoring to Hedera mainnet requires the RUBRIC_API_KEY environment variable. Desktop MCP hosts only; no web executable.
Installation
In SourceWeft
- Open Rubric Protocol in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
@rubric-protocol/mcp-server
Post-quantum AI compliance attestation as an MCP server. Works with Claude Desktop, Claude Code, Cursor, Windsurf, Continue, Zed, and any MCP-compatible host.
v2.0 — now with free local tier. No API key required to start attesting.
Install
Configure (Claude Desktop)
Edit claude_desktop_config.json:
Restart Claude Desktop. Rubric tools appear in the MCP menu.
Tiers
Free (no key)
attest— PQ-signed Merkle leaves stored in~/.rubric/local-bundles/verify— check local attestationsframework_detect— detect applicable regulations offlinecost_estimate— project monthly costregister_agent— request a free key to unlock HCS anchoringstatus— federation health
With API key (free developer tier)
Everything above, plus:
- HCS anchoring on Hedera mainnet (tamper-evident third-party timestamp)
get_proof— ZK Merkle inclusion proofs (Noir/Barretenberg)
Request a key via the register_agent tool, or at https://rubric-protocol.com
Standard+ tier ($999/mo)
compliance_tag— attach regulatory framework tags to bundlesbundle_query— filter by leafType, agentId, time range- 100K attestations/mo, overage $0.01 each
Enterprise ($9,999/mo) / Dedicated ($25K+/mo)
Higher throughput, SLA, dedicated federation capacity.
Paid evidence tools (x402) - new in 2.2
Six tools that pay per call in USDC on Base via the x402 protocol. No Rubric account or API key required - just a funded wallet. Every response carries a signed, Hedera-anchored attestation ID your agent can cite later.
Setup (3 steps)
- Create a wallet and fund it with a few dollars of USDC on Base (Coinbase -> withdraw USDC -> network: Base)
- Set
RUBRIC_WALLET_KEYto the wallet's private key in your MCP server environment - Optional:
RUBRIC_X402_DAILY_LIMIT(default1.00USD/day)
Money safety, by design
- No wallet key? Paid tools return setup guidance - never errors, never charges.
- Daily ceiling. Spending stops at your limit; the tool returns a budget error your agent can read. Resets 00:00 UTC.
- Price protection. Before paying, each tool checks the server's quoted price against its documented maximum and refuses anything higher - even we cannot overcharge you.
- Full accounting. Every paid response includes
spentTodayUsd. - Your key never leaves the MCP process. A failed operation is never charged.
Catalog, prices, and terms are machine-readable at x402.json and openapi.json.
Module configuration
Tools load by module via RUBRIC_MCP_MODULES (default: core,x402).
Available: core, x402, attestation, verification, compliance,
regulatory, governance, registry, ops, or all.
Regulatory coverage
EU AI Act (Articles 9–15, 17, 26, 49, 72, 73, Annexes IV/XI/XII), SR 11-7, OCC, FDIC, NIST AI RMF 1.0, TX TRAIGA, CO AI Act, HIPAA, EU DSA, NIS2, SEC, CFTC, ECOA Reg B, NYC LL144.
Tools (50+ across 9 modules)
attest— attest an AI decision (local free, HCS with key)verify— verify attestation by IDget_proof— ZK Merkle inclusion proofregister_agent— request free developer API keystatus— federation healthframework_detect— auto-detect regulatory frameworkscost_estimate— monthly cost projectioncompliance_tag— attach framework tags (Standard+)bundle_query— query bundles (Standard+)
Architecture
- Post-quantum signatures: ML-DSA-65 (NIST FIPS 204) via liboqs
- Merkle aggregation: N-tier, SHA3-256, up to 1,000,000:1 compression
- Anchoring: Hedera mainnet HCS topic 0.0.10416909
- Federation: 5 geo-distributed nodes (US, SG, JP, CA, EU)
- ZK proofs: Noir beta.19, depth-20 Poseidon2 Merkle inclusion
Links
- Homepage: https://rubric-protocol.com
- Pricing: https://rubric-protocol.com/pricing
- Status: https://rubric-protocol.com/status
- Changelog: ./CHANGELOG.md
Source: README.md at commit 882d257
Tools
0Version history
1- v2.0.2LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts