π‘ Why ProofGraph?
Traditional Retrieval-Augmented Generation (RAG) dumps 5,000 to 20,000 unverified tokens of noisy HTML, sidebars, and duplicate paragraphs into LLM prompts. This causes context bloat, higher token costs, latency spikes, and hallucination amplification.
ProofGraph replaces fuzzy text dumping with Minimum Sufficient Context backed by Cryptographic Provenance:
Question βββΊ Entity Resolution βββΊ Claim Verification βββΊ Evidence Ranking βββΊ Token Optimizer βββΊ Compact Proof Packet
π― ProofGraph retrieves proof, not just text.
β¨ Key Features
| Feature | Description | Benefit |
|---|
| π‘οΈ Evidence-First Verification | Cryptographic SHA-256 hashes & timestamped citations | Eliminates unsupported hallucinations |
| π Multi-Signal Entity Resolution | Combines Jaro-Winkler, domain, GitHub org & npm scope | Resolves aliases without false-positive merges |
| β‘ Token Budget Optimizer | Enforces strict budget caps (500β5,000 tokens) | ~90.7% context reduction |
| πͺ’ Contradiction Detection | Detects and exposes conflicting sources transparently | Neutral, unbiased dispute analysis |
| π 12 Native MCP Tools | Tools for packets, traversal, audits & Schema.org JSON-LD | Instant drop-in for Claude, Antigravity, Cursor |
| π Enterprise SSRF Protection | Blocks loopback, RFC 1918 subnets, and cloud metadata | Safe for automated agentic execution |
π Quick Start
1. Run Instantly with npx
No installation required:
npx -y @aibuildinfra/proofgraph
2. Install Globally via npm
npm install -g @aibuildinfra/proofgraph
π€ MCP Client Configuration
Claude Desktop / Google Antigravity / Cursor
Add ProofGraph to your MCP configuration file (claude_desktop_config.json or Antigravity/Cursor MCP settings):
{
"mcpServers": {
"proofgraph": {
"command": "npx",
"args": ["-y", "@aibuildinfra/proofgraph"]
}
}
}
π οΈ MCP Tools Suite
ProofGraph equips AI agents with 12 powerful verification and retrieval tools:
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β PROOFGRAPH MCP TOOLS β
βββββββββββββββββββββββββββ¬βββββββββββββββββββββββββββββββββββββββββββββββββββ€
β get_evidence_packet β Primary retrieval: Minimal sufficient proof β
β search_entities β Multi-signal fuzzy & alias entity matching β
β get_entity β Canonical record, relationships & Schema.org β
β verify_claim β Verify assertions against cryptographic proof β
β find_evidence β Token-budgeted evidence extraction for a claim β
β find_relationships β 1-hop, 2-hop, 3-hop graph traversal paths β
β explain_entity β Concise factual entity profile & key edges β
β trace_claim β Step-by-step provenance audit trail β
β compare_claims β Impartial discrepancy detection across claims β
β analyze_web_presence β Digital footprint analysis across 8 platforms β
β audit_entity_consistencyβ Detect mismatches in names, domains & metadata β
β export_graph β Export graph to JSON, JSON-LD, CSV, GraphML β
βββββββββββββββββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββββββββββββ
Example: Evidence-First Response Packet
When an AI agent queries:
{
"question": "Is AI Build Infra associated with HumanCraft and ProofGraph?"
}
ProofGraph returns a compact, citation-grounded evidence packet:
{
"entities": [
{
"id": "entity:organization:ai-build-infra",
"name": "AI Build Infra",
"canonical_url": "https://aibuildinfra.com/"
}
],
"claims": [
{
"claim": "AI Build Infra develops MCP servers and open-source agent tooling.",
"status": "supported",
"confidence": 0.98
}
],
"relationships": [
{ "source_id": "entity:organization:ai-build-infra", "relationship": "DEVELOPS", "target_id": "entity:project:proofgraph" },
{ "source_id": "entity:organization:ai-build-infra", "relationship": "DEVELOPS", "target_id": "entity:project:humancraft" }
],
"evidence": [
{
"source": "MCP Registry",
"url": "https://registry.modelcontextprotocol.io/servers/io.github.AI-BuildInfra/humancraft-ui",
"excerpt": "Server identifier io.github.AI-BuildInfra/humancraft-ui registered under AI Build Infra publisher identity.",
"confidence": 0.99
}
],
"provenance": [
{
"source_id": "source:registry:mcp-official",
"content_hash": "10c56a410a5b22e6f3037126be38073b20e950c2ded2ee36815f20072041b889",
"retrieved_at": "2026-09-25T12:00:00Z"
}
],
"metrics": {
"raw_estimated_tokens": 4500,
"returned_tokens": 485,
"compression_ratio": 0.108
}
}
π Token Reduction Benchmark
ProofGraph dramatically cuts LLM context consumption without sacrificing retrieval precision:
| Benchmark Query | Traditional Raw RAG | ProofGraph Packet | Token Savings | Verified Citations |
|---|
| "Who is AI Build Infra?" | ~4,500 tokens | 999 tokens | 77.8% | 4 primary proofs |
| "What is HumanCraft?" | ~4,500 tokens | 219 tokens | 95.1% | Grounded profile |
| "What evidence connects HumanCraft to AI Build Infra?" | ~4,500 tokens | 221 tokens | 95.1% | Relationship path |
| "Is AI Build Infra associated with ProofGraph?" | ~4,500 tokens | 485 tokens | 89.2% | Direct proof + rels |
| "What services does AI Build Infra provide?" | ~4,500 tokens | 164 tokens | 96.4% | Grounded service claim |
| Cumulative Total | ~22,500 tokens | 2,088 tokens | 90.7% Context Reduction | 100% Traceable |
Run the benchmark locally anytime:
ποΈ Architecture
flowchart TD
Agent[AI Agent / Claude / Antigravity / Cursor] <--> MCP[ProofGraph MCP Protocol Layer]
subgraph Engine [Deterministic Verification Engine]
MCP <--> Optimizer[Token Budget & Deduplication Engine]
MCP <--> Resolver[Multi-Signal Entity Resolver]
MCP <--> Verifier[Claim & Contradiction Verifier]
MCP <--> Traversal[1/2/3 Hop Graph Traversal]
end
subgraph Storage [Cryptographic Provenance Store]
Entities[(Canonical Entities)]
Claims[(Verified Claims)]
Evidence[(Evidence & SHA-256 Hashes)]
Sources[(Authoritative Sources)]
Relationships[(Typed Graph Multigraph)]
end
Engine <--> Storage
subgraph Security [Security & Safety Boundary]
SafeFetch[SSRF Protection\nBlocks RFC1918 & Cloud Metadata]
Robots[Robots.txt & Rate Limiting]
end
Engine <--> Security
π» Developer CLI & Web Dashboard
CLI Commands
# Initialize seed graph
proofgraph init
# Search entities with multi-signal matching
proofgraph search "AI Build Infra"
# Verify a claim with grounded evidence
proofgraph verify "AI Build Infra develops MCP servers"
# Display full cryptographic provenance trail
proofgraph trace "AI Build Infra develops MCP servers"
# Run cross-platform consistency audit
proofgraph audit "AI Build Infra"
# Run AI discoverability checklist
proofgraph discoverability "AI Build Infra"
# Export graph to JSON-LD / CSV / GraphML
proofgraph export jsonld
Interactive Web Dashboard
Explore graph topologies, entities, claims, and audit diagnostics visually:
npm run dashboard
# Open http://localhost:3456
π Security & Ethical Guidelines
ProofGraph is built under strict open-source safety principles:
- SSRF Prevention: All outbound HTTP requests block private subnets (
127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16), link-local IPs, and Cloud Metadata (169.254.169.254, metadata.google.internal).
- Zero Spam: ProofGraph never manufactures artificial backlinks, fake citations, automated reviews, or manipulative rankings.
- Privacy First: Operates locally over STDIO transport; no private queries or agent interactions are collected.
π Technical Documentation