proofrelay
A read-only MCP verifier that helps agents and operators manage action checkpoints, inspect evidence bundle metadata, and flag risk signals before relying on external tools or services.
Get Verifier Status (
proofrelay.get_verifier_status): Retrieve public ProofRelay discovery metadata (hosted MCP endpoint URL, server-card URL, product URL, available tool count, trust boundary notes) — no authentication required.Recommend Checkpoint (
proofrelay.recommend_checkpoint): Determine whether an agent action class (e.g.,paid_tool_call,revenue_action,financial_transaction,external_reliance) requires a checkpoint before proceeding. Returnspass,skipped, orneeds_reviewwith a reason.Verify Bundle (
proofrelay.verify_bundle): Check the shape and consistency of synthetic or non-confidential evidence bundle metadata for local pass/fail boundary checks. Does not certify real-world facts and must not receive secrets or sensitive data.Scan MCP Risk (
proofrelay.scan_mcp_risk): Inspect caller-supplied public MCP descriptor metadata for advisory risk signals (e.g., mutating tools, payment/wallet language, credential indicators, missing public metadata) — without fetching the server URL or acting as a security certification.Describe CLI/SDK Helper (
proofrelay.describe_cli_sdk_helper): Get a read-only integration guide explaining how to wrap agent or CLI actions with ProofRelay-style pass/fail/review checkpoint patterns.
GENESIS ProofRelay MCP
ARCHIVED NON-CANONICAL SNAPSHOT. This repository is retained for public history and marketplace-link continuity. It is not the canonical ProofRelay implementation, is not a source mirror, and is not suitable for production reliance. The executable local wrapper formerly stored here intentionally implemented only shallow discovery checks and has been removed to prevent it from being confused with the hosted verifier.
Documentation-only discovery snapshot for the GENESIS ProofRelay MCP verifier. The canonical public machine-readable contract is the live server card.
ProofRelay's hosted public surface is a read-only MCP verifier for synthetic or
non-confidential evidence bundle metadata. This repository does not implement
or reproduce that verifier.
ProofRelay is not an oracle, escrow service, DeFi risk engine, legal opinion, or endorsement system. It does not certify real-world truth, prove title, approve payments, custody funds, or make a third-party tool safe.
Why This Exists
Agents increasingly take actions that cross from "answer a question" into "rely on something external": paid tool calls, revenue actions, financial transactions, closing workflows, marketplace listings, and MCP integrations. Those actions need simple controls that work before private data or production systems are exposed.
ProofRelay gives agents a small decision layer:
decide whether a checkpoint is needed
check whether public-safe evidence metadata is shaped consistently
surface mismatch or missing-authority signals
keep secret-bearing material out of public verification paths
record pass/fail/review outcomes without claiming real-world certification
Related MCP server: Hive Radar MCP
Snapshot status
Archived from public repository commit
feeca01f31fead61a77cb1be9691dd65eec596a0on 2026-08-27.Hosted surface observed on 2026-08-27: 26 tools, 18 resources, 13 prompts.
Those counts are observations, not a frozen API guarantee. Read the live server card before every integration or reliance decision.
Security review of the prior 17-commit history found no secret with Gitleaks; this is not a guarantee that prior clones or third-party mirrors are current.
Reopening requires the controls in
export-policy.json, protected review, and an explicitly assigned maintenance owner.
Decision Vocabulary
The hosted service uses these high-level statuses:
Status | Meaning | Typical next step |
| The public-safe metadata satisfies the local check. | Continue, while preserving the evidence reference. |
| The action or bundle has missing, ambiguous, or higher-reliance signals. | Pause for operator, policy, or stronger evidence review. |
| The action class is low-reliance or no checkpoint applies. | Continue without treating the result as verified. |
| The bundle shape or required metadata is inconsistent. | Do not rely on the bundle until corrected. |
Public Tool Surface
Tool names, schemas, counts, and descriptions change independently of this archived snapshot. Discover them only from the live server card and MCP protocol. Do not infer hosted behavior from this repository's Git history.
ProofRelay vs DeFi Safety
ProofRelay is about evidence and action-control metadata. DeFi safety tooling is about smart-contract, protocol, wallet, liquidity, oracle, bridge, and token risk. Those are different domains.
ProofRelay may help an agent notice that a financial or wallet-related action needs review, or that a bundle lacks a required authority envelope. It does not audit contracts, verify liquidity, price assets, validate token economics, score protocol risk, or make custody decisions.
Public Links
Glama listing: https://glama.ai/mcp/servers/GENESISRE/proofrelay
Smithery listing: https://smithery.ai/servers/genesis/proof-relay
Product page: https://genesisre.io/proofrelay
MCP endpoint: https://mcp.genesisre.io/mcp
MCP health: https://mcp.genesisre.io/health
MCP server card: https://mcp.genesisre.io/.well-known/mcp/server-card.json
Agent service descriptor: https://genesisre.io/.well-known/proofrelay-agent-service.json
Install From Smithery
npx -y smithery mcp add genesis/proof-relay
npx -y smithery mcp get proof-relay
npx -y smithery tool list proof-relay
npx -y smithery tool call proof-relay proofrelay.get_verifier_status '{}'Generic MCP Client
{
"mcpServers": {
"genesis-proofrelay": {
"type": "http",
"url": "https://mcp.genesisre.io/mcp"
}
}
}Trust Boundary
ProofRelay verifies submitted synthetic or non-confidential evidence bundle shape and integrity signals. It does not process payments, custody funds, escrow assets, certify legal/title/compliance status, certify real-world facts, provide legal advice, review private source code, or require source-code disclosure.
Do not submit secrets, private prompts, raw logs, source code, customer files, wallet keys, payment credentials, or tenant traces to the public MCP endpoint.
Public Repository Boundary
This archived repository is intentionally limited to public ProofRelay MCP
discovery, setup, and trust-boundary documentation. The hosted MCP service is
operated separately by GENESIS at mcp.genesisre.io.
This repository does not include:
internal GENESIS monorepo code
production deployment automation
private credentials or environment files
paid settlement adapters
Skyfire, Coinbase, Stripe, AWS, or internal operator secrets
customer data, raw logs, prompts, traces, or evidence bundles
a local MCP server, SDK, verifier, container, or deployable package
See PUBLICATION_BOUNDARY.md for the publication allowlist and exclusion policy.
License
MIT. See LICENSE.
Available Tools
5 toolsproofrelay.describe_cli_sdk_helperARead-onlyIdempotent
Read-only integration guide for the invisible pass/fail/review helper pattern. Use when an agent or developer needs to understand how to wrap actions with ProofRelay-style checkpoint statuses; use recommend_checkpoint for a specific action decision. Requires no authentication, mutates nothing, and returns JSON text describing statuses, hosted endpoint, and the action-to-audit flow.
Inputs: no named parameters. Returns: helper, statuses, hosted_mcp_endpoint, example_flow, +1 more
proofrelay.get_verifier_statusARead-onlyIdempotent
Read-only status lookup for public ProofRelay discovery metadata. Use first to confirm the hosted MCP endpoint, server-card URL, product URL, public tool count, and trust boundary. Requires no authentication, performs no network call from this local wrapper, mutates nothing, and returns JSON text with status, URLs, counts, and non-confidential boundary notes.
Inputs: no named parameters. Returns: status, server_name, hosted_mcp_endpoint, server_card, +4 more
proofrelay.recommend_checkpointARead-onlyIdempotent
Read-only checkpoint recommendation for a high-level agent action class. Use before relied-upon actions such as paid_tool_call, revenue_action, external_reliance, or financial_transaction; use get_verifier_status for discovery only and verify_bundle after evidence metadata exists. Requires no authentication, mutates nothing, and returns JSON text with status pass, skipped, or needs_review plus checkpoint and reason fields.
Inputs: action_class (required), has_authority_envelope. Returns: status, checkpoint, reason, action_class, +1 more
proofrelay.scan_mcp_riskARead-onlyIdempotent
Read-only advisory scan of caller-supplied public MCP descriptor metadata. Use before registering or trusting an MCP server to flag mutating tools, payment or wallet language, credential signals, or missing public metadata; use verify_bundle for evidence bundles instead. Requires no authentication, does not fetch server_url, mutates nothing, is not a security certification, and returns JSON text with status, findings, recommended_control, non_claims, and boundary fields.
Inputs: server_url, descriptor. Returns: status, findings, recommended_control, non_claims, +1 more
proofrelay.verify_bundleARead-onlyIdempotent
Read-only public-safe bundle shape check for synthetic or non-confidential evidence metadata. Use when you already have a bundle fixture and need a local pass/fail boundary check before sharing or routing to hosted verification; use recommend_checkpoint when deciding whether a checkpoint is needed. Requires no authentication, mutates nothing, does not certify real-world facts, and returns JSON text with status, checks, problems, verifier, and hosted endpoint fields.
Inputs: bundle (required). Returns: status, verifier, checks, problems, +3 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 5 tools
Each tool targets a clearly distinct concern: discovery status, checkpoint recommendation, bundle verification, MCP risk scanning, and helper documentation. The descriptions cross-reference one another and specify exactly when each tool should be used, leaving little room for misselection.
All tool names follow a consistent lowercase snake_case verb_object pattern: get_verifier_status, recommend_checkpoint, verify_bundle, scan_mcp_risk, describe_cli_sdk_helper. The verbs and objects are specific and predictable, with no mixed conventions or vague generic names.
With exactly 5 tools, the server is well-scoped for a read-only advisory utility. Each tool earns its place by covering a separate step in the ProofRelay workflow without redundancy or unnecessary expansion.
For the server's declared read-only advisory purpose, the tool surface is complete: discovery status, checkpoint recommendation, bundle verification, MCP risk scanning, and integration guidance are all present. The descriptions clearly separate local checks from hosted verification, so there are no obvious dead ends within the intended scope.
Maintenance
Related MCP Connectors
Read-only verifier for 25 ProofRelay MCP tools and non-confidential evidence bundles.
Read-only Remote MCP for externally grounded AI agent trust receipts.
Read-only MCP: free OpenAI security evidence ledger (55 fields) + SaaSDossier release register.
Provenance and metadata tools for AI agents. Paid tools use x402 (USDC) at /mcp/<tool>.
Related MCP Servers
- AlicenseAqualityDmaintenanceSafety layer for autonomous DeFi agents. Scans contracts for exploit patterns, simulates transactions, blocks honeypots.411 npm1MIT
- AlicenseNot gradedqualityBmaintenanceProvides read-only MCP tools to list archived snapshots, retrieve methodology and proof bundles, and verify supplied evidence bundles.MIT
- AlicenseNot gradedqualityCmaintenanceEnables proof-carrying project execution through a read-only remote MCP endpoint and WebMCP tools for managing Project Passports, task graphs, evidence-backed provider comparisons, route approvals, artifacts, verification records, and checkpoints without executing external providers.MIT
- AlicenseNot gradedqualityBmaintenanceEnables free, read-only checks of AI agents and MCP servers for conformance, published changes, delivery evidence, and verifiable receipts without requiring an account or API key, using public or non-personal data only.MIT