@attested-intelligence/aga-mcp-server 0.1.1 → 2.0.0

package/dist/resources/specification.js

@@ -0,0 +1,107 @@
+export const PROTOCOL_SPECIFICATION = `# Attested Governance Artifact (AGA) Protocol Specification v2.0.0
+
+## Patent Reference
+USPTO Application No. 19/433,835
+
+## NIST References
+- NIST-2025-0035: AI Agent Transparency and Accountability
+- NCCoE AI Agent Identity and Authorization
+
+## Protocol Overview
+The AGA protocol provides cryptographic governance for autonomous AI systems through:
+1. **Sealed Hash Attestation** - SHA-256(bytes_hash || metadata_hash || policy_ref || seal_salt)
+2. **Continuity Chain** - Tamper-evident append-only event log with privacy-preserving leaf hashes
+3. **Portal State Machine** - Zero-trust Policy Enforcement Point (7 states, fail-closed)
+4. **Signed Receipts** - Ed25519-signed measurement receipt for EVERY measurement
+5. **Evidence Bundles** - Offline-verifiable packages with Merkle inclusion proofs
+
+## 10 Measurement Embodiments
+1. EXECUTABLE_IMAGE - Runtime binary or script content
+2. LOADED_MODULES - Dynamic libraries and plugins
+3. CONTAINER_IMAGE - Container image manifest hash
+4. CONFIG_MANIFEST - Configuration file integrity
+5. SBOM - Software Bill of Materials verification
+6. TEE_QUOTE - Trusted Execution Environment attestation
+7. MEMORY_REGIONS - Runtime memory layout verification
+8. CONTROL_FLOW - Execution path integrity
+9. FILE_SYSTEM_STATE - Filesystem integrity monitoring
+10. NETWORK_CONFIG - Network configuration baseline
+
+## 6 Portal States
+1. INITIALIZATION - Server started, no artifact loaded
+2. ARTIFACT_VERIFICATION - Verifying artifact signature and validity
+3. ACTIVE_MONITORING - Operational, measurements occurring
+4. DRIFT_DETECTED - Hash mismatch detected, enforcement pending
+5. PHANTOM_QUARANTINE - Forensic capture mode, outputs severed
+6. TERMINATED - Fail-closed, no recovery without re-attestation
+
+Plus SAFE_STATE for graceful degradation on revocation.
+
+## 7 Enforcement Actions
+1. QUARANTINE - Phantom execution with forensic capture
+2. TERMINATE - Immediate kill, fail-closed
+3. SAFE_STATE - Return-to-home / controlled shutdown
+4. NETWORK_ISOLATE - Sever network, continue local
+5. KEY_REVOKE - Invalidate cryptographic keys
+6. TOKEN_INVALIDATE - Revoke access tokens
+7. ALERT_ONLY - Log without enforcement (gradual deployment)
+
+## 3 Verification Tiers
+| Tier | Description | Trust Assumption |
+|------|-------------|-----------------|
+| Bronze | Cryptographic signatures only | Trust signing keys |
+| Silver | Signatures + continuity chain | Trust chain operator + keys |
+| Gold | Full verification with blockchain-anchored Merkle proofs | Minimal trust - external anchor |
+
+## 3 Disclosure Modes
+1. PROOF_ONLY - Returns boolean attestation without revealing the value
+2. REVEAL_MIN - Returns minimal representation (e.g., range instead of exact value)
+3. REVEAL_FULL - Returns the complete claim value
+
+## Leaf Hash Formula (Claim 3c - Privacy Innovation)
+\`\`\`
+leaf_hash = SHA-256(
+  schema_version || "||" || protocol_version || "||" ||
+  event_type || "||" || event_id || "||" ||
+  sequence_number || "||" || timestamp || "||" ||
+  previous_leaf_hash
+)
+\`\`\`
+**PAYLOAD IS EXCLUDED from the leaf hash.** This is the key patent innovation - chain integrity can be verified without revealing the contents of any event. Only the structural metadata participates in the hash. The payload is separately integrity-protected via event_signature.
+
+## SPIFFE/SPIRE Integration Point
+SPIRE handles node-to-workload identity (SVID); AGA handles workload-to-intent governance. SPIFFE provides transport-layer identity binding via SVIDs (SPIFFE Verifiable Identity Documents). AGA binds governance policy to the workload's operational intent, creating a complementary layer:
+- SPIFFE: "This workload IS who it claims to be" (identity)
+- AGA: "This workload IS DOING what it was attested to do" (governance)
+
+## Framework Alignment
+| Framework | AGA Alignment |
+|-----------|--------------|
+| NIST SP 800-53 | SI-7 (Software Integrity), AU-10 (Non-repudiation), SI-4 (Monitoring) |
+| NIST AI RMF | Govern → Policy Artifacts; Map → Subject ID; Measure → Portal + Receipts; Manage → Enforcement |
+| NIST SP 800-57 | Key management for Ed25519 signing keys |
+| NIST SSDF (SP 800-218) | Software supply chain integrity via sealed hash attestation |
+| NIST SP 800-207 (ZTA) | Zero Trust Architecture - portal as Policy Enforcement Point, never trust, always verify |
+| ISO 42001 | AI management system - governance artifacts as compliance evidence |
+| EU AI Act | High-risk AI system transparency via evidence bundles |
+
+## Cryptographic Primitives
+- **Hashing:** SHA-256 (primary), BLAKE2b-256 (secondary)
+- **Signing:** Ed25519 via @noble/ed25519
+- **Salts:** 128-bit CSPRNG via @noble/hashes/utils
+- **Merkle Trees:** SHA-256 binary tree with inclusion proofs
+- **Serialization:** RFC 8785 deterministic JSON (sorted keys)
+
+## Event Types (12)
+GENESIS, POLICY_ISSUANCE, INTERACTION_RECEIPT, REVOCATION, ATTESTATION,
+ANCHOR_BATCH, DISCLOSURE, SUBSTITUTION, KEY_ROTATION, BEHAVIORAL_DRIFT,
+DELEGATION, RE_ATTESTATION
+
+## 4 Sensitivity Levels
+- S1_LOW - Can be revealed fully
+- S2_MODERATE - Can be revealed minimally or proved
+- S3_HIGH - Proof only, auto-substitutes to lower sensitivity
+- S4_CRITICAL - Maximum protection, proof only, cascading substitution
+`;
+export const SPECIFICATION_URI = 'aga://specification';
+//# sourceMappingURL=specification.js.map

package/dist/resources/specification.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"specification.js","sourceRoot":"","sources":["../../src/resources/specification.ts"],"names":[],"mappings":"AAAA,MAAM,CAAC,MAAM,sBAAsB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAwGrC,CAAC;AAEF,MAAM,CAAC,MAAM,iBAAiB,GAAG,qBAAqB,CAAC"}

package/dist/server.d.ts

@@ -1,12 +1,9 @@
 /**
- * AGA MCP Server. The Portal (ref 150) as an MCP service.
+ * AGA MCP Server V2.0.0 - The Portal (ref 150) as an MCP service.
  *
- * V3 NIST-aligned behaviors:
- * 1. Every measurement generates a receipt (match OR mismatch)
- * 2. TTL checked on every measurement (fail-closed)
- * 3. Mid-session revocation via revoke_artifact tool
- * 4. Governance middleware: portal state checked before tool execution
- * 5. Auto-chaining: every operation writes to continuity chain
+ * 20 tools, 4 resources, 3 prompts.
+ * USPTO Application No. 19/433,835
+ * NIST-2025-0035, NCCoE AI Agent Identity and Authorization
  */
 import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 export declare function createAGAServer(): Promise<McpServer>;

package/dist/server.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"server.d.ts","sourceRoot":"","sources":["../src/server.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AACH,OAAO,EAAE,SAAS,EAAE,MAAM,yCAAyC,CAAC;AAiDpE,wBAAsB,eAAe,IAAI,OAAO,CAAC,SAAS,CAAC,CAwX1D"}
+{"version":3,"file":"server.d.ts","sourceRoot":"","sources":["../src/server.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAAE,SAAS,EAAE,MAAM,yCAAyC,CAAC;AAwCpE,wBAAsB,eAAe,IAAI,OAAO,CAAC,SAAS,CAAC,CAwW1D"}