audiencescore
OfficialAudienceScore
AudienceScore is an open protocol for review scores gated by cryptographic proof of a real transaction or verified participation. A signed receipt unlocks one review for one versioned offering, and every published score is a signed, recomputable rendering over the ledger.
Status
AudienceScore is a live pilot deployment, pre-cryptographic-audit, not a
production issuance system. Pilot receipts, events, and score manifests carry
env: "pilot" in signed bodies; the pilot ledger may be reset and receipts may
be re-issued after audit.
Two gates remain open before non-pilot issuance:
Independent cryptographic review of the receipt and rendering-signature scheme.
Per-vertical legal review before any regulated profile ships.
Related MCP server: agentforge-trust-mcp
Connect MCP
Add this remote MCP server URL in any client that supports Streamable HTTP:
https://mcp.audiencescore.org/mcpNo account or API key is required. The pilot server exposes read-only
get_score and get_score_evidence tools for v0.2 rendering manifests.
Read a Score
Fetch a signed pilot score manifest with one curl:
curl -sS https://mcp.audiencescore.org/v0/scores/field-elevate-demo%40v1Below the k-anonymity floor, the manifest returns published: false rather
than a fabricated score.
Verify that the manifest signature is valid and that its signer is in the published key set:
node - <<'NODE'
const { verifyPayload } = require('./reference-impl/src/crypto');
const scoreUrl = 'https://mcp.audiencescore.org/v0/scores/field-elevate-demo%40v1';
const keysUrl = 'https://audiencescore.org/.well-known/audiencescore-keys.json';
(async () => {
const signed = await fetch(scoreUrl).then((r) => r.json());
const keys = await fetch(keysUrl).then((r) => r.json());
const publishedKeys = new Set((keys.keys || []).map((k) => k.key));
const inKeySet = publishedKeys.has(signed.signer);
const valid = verifyPayload(signed.signer, signed.manifest, signed.sig);
console.log(JSON.stringify({ published: signed.manifest.published, inKeySet, valid }, null, 2));
if (!inKeySet || !valid) process.exit(1);
})();
NODERun Locally
Requires Node.js 24+.
git clone https://github.com/audiencescore/audiencescore.git
cd audiencescore
npm install --prefix reference-impl
npm test --prefix reference-impl
node reference-impl/demo.jsThe demo proves the full current loop: receipt issued, review admitted, score rendered, evidence recomputed, and score queried over MCP.
Spec and Governance
Spec v0.2a: receipts, roles, attestation levels, versioned offerings, invariants, and threat model.
Rendering v1: current score math.
Conformance: signed vectors, canonical bytes, and Python verifier.
Acceptance tests: AT-1 through AT-25, all executable in CI.
Pilot docs: issuer, review, score-reading, and deployment runbooks.
Governance: open score math, sealed-admission boundaries, and accountability rules.
DRIFT.md: the plain-English record of v0.1 to v0.2 drift decisions.
Contribute
Protocol changes start with the spec. Anything touching receipts, invariants, rendering math, or conformance vectors needs a spec/RFC discussion before code. Implementation changes need tests, DCO sign-off, and passing CI.
Start with CONTRIBUTING.md. Security reports go through SECURITY.md.
This server cannot be deployed
Maintenance
Related MCP Connectors
Scan any website or MCP server for agent-trust-readiness; returns a signed, verifiable scorecard.
Experimental MCP server for current empirical verification of explicit public HTTPS endpoint claims.
Read-only MCP server that answers questions about turva.dev from its published data. Five tools return JSON: the service catalog with prices, contact and operator details, engagement principles and dated agent-readiness and security evidence with verification links. Connect over Streamable HTTP. You need no API key. The server does not scan other websites or run audits.
MCP server providing access to the Scorecard API to evaluate and optimize LLM systems.
Related MCP Servers
- AlicenseAqualityDmaintenanceMCP server for AI agent identity — verify agents with Ed25519 signatures, check trust scores, sign and verify content, exchange encrypted messages. Built on the Agent Identity Protocol (AIP).8MIT
- AlicenseAqualityDmaintenanceMCP server exposing AgentForge Trust Score audit tools. Query trust, evaluate policies, list trusted servers, recommend with filter.419 npm1MIT
- AlicenseAqualityFmaintenanceA thin MCP server that wraps provenance-receipts and quality-gate services as tools, enabling AI agents to certify content origin and score quality via Ed25519-signed receipts.541 npmMIT
- AlicenseAqualityDmaintenanceIndependent trust scores, claim audits, and side-by-side comparisons for AI agents, queryable over MCP. Every verdict is backed by hands-on testing and signed evidence from Hlido (hlido.eu). Hosted endpoint available at https://hlido.eu/mcp — no auth required.14MIT