Agent Provenance GraphFree ForeverDeterministic

HyperStack: The Agent Provenance Graph
for Verifiable AI

Build agents you can audit and trust at $0/operation. Timestamped facts, auditable decisions, deterministic trust propagation, and ledger-like accountability eliminating black-box chaos in multi-agent systems.

t₁t₂t₂t₃
50 cards free · No credit card · 30s setup
📦 npm · hyperstack-core ⭐ GitHub · hyperstack-core 🦞 ClawHub · hyperstack
$0 per operation
Full audit replay
30s setup
Deterministic trust
4 SDKs live
Provenance-Ready Integrations
🔗 OpenClaw⚡ Claude Code🔗 Cursor🤖 LangChain👥 CrewAI🐍 Python💚 Node.js⚙️ VS Code🌐 REST API🔗 LangGraph 🔗 OpenClaw⚡ Claude Code🔗 Cursor🤖 LangChain👥 CrewAI🐍 Python💚 Node.js⚙️ VS Code🌐 REST API🔗 LangGraph
PROVENANCE OVER CHAOS

Prove chain-of-custody for agent knowledge.
Reconstruct exactly what an agent knew at decision time.

Can't trust agent decisions? HyperStack stamps provenance for deterministic resolution. Detect hindsight changes. No black-box guessing.

❌ Black-box agents
# "Why did the agent decide X?" # → No audit trail # → Hindsight changes invisible # "What blocks deploy?" # → Fuzzy RAG, no typed relations
✔ Agent Provenance Graph
await hs.replay("decision-use-clerk"); // → Exact state at decision time // → Detect hindsight changes await hs.blockers("deploy-prod"); // → [migration-23, staging-tests]

Timestamped facts. Ledger-like accountability. Build agents you can audit and trust at $0/operation.

PROVENANCE-FIRST FEATURES

Verifiable agents. Ledger-like accountability.

Build agents you can audit and trust at $0/operation.

⚙️

Provable Schemas

Developer-defined types ensure no hallucinations, with zero-cost exact queries. Your schema is your contract no LLM guessing.

⏳

Decision Replay

Full graph versioning for auditing what agents knew and why decisions were made. Reconstruct exact state at any point in time.

🔗

Semantic Graph

Typed edges (blocks, depends-on, owns, decided) for real-time traversal of blockers and impacts. Ask "what blocks deploy?" get exact answers.

🛡️

Trust Layer

Agent provenance stamps and scores for deterministic conflict resolution without LLMs. Know who stored what, when.

// Resolve conflicts deterministically
hs.resolveConflict(slug, { trustThreshold: 0.8 })
🧲

GraphRAG

Hybrid vector + graph search, provenance-aware. Semantic similarity finds seed cards. Graph traversal surfaces everything connected with full audit trail.

🧭

Agentic Routing

Ask in plain English. "What breaks if proj-auth changes?" – HyperStack detects intent, picks impact mode. Provenance-enhanced provability.

🌿

Audit Branches

Fork for experiments, diff/merge with full accountability. Git for agent provenance parent graph always untouched.

🔍

Provenance Queries. One API.

Blockers, impact, replay, and graph traversal all provenance-aware. Reconstruct what agents knew. Detect hindsight changes. Full audit trail.

// Blockers – what prevents this? await hs.blockers("deploy-prod"); // Impact – blast radius of a change await hs.impact("use-clerk"); // Decision Replay – what did we know then? await hs.graph("auth-api", { at: "2026-02-01" });
COMPARISON

Provenance tracking others can't match.

HyperStackMem0BackboardCognee
Provenance TrackingTimestamped, replayable factsNo timestamped replayThread-onlyNo decision replay
Trust ResolutionDeterministic, no LLMLLM-dependentLLM for resolutionLLM-dependent
Audit ReplayFull graph at any timestamp
Typed relationsblocks, depends-on, owns, decidedAuto-extractedThreads onlyGeneric graph
Cost per op$0 deterministic~$0.002 LLMUsage-based~$0.002 LLM
GraphRAGVector + graph, provenance-awareVector only Partial
Branching / DiffFork, merge, full accountability
Multi-Agent CoordinationDeterministic trust propagationShared userId issuesThread-basedLLM-dependent / None
Setup30 seconds5-10min + Docker5 min5+ min + Neo4j
PROVENANCE-READY INTEGRATIONS

Drop into your stack in 30 seconds.

LangGraph, Cursor, OpenClaw with audit trails.

// Claude Desktop / Cursor / VS Code npx hyperstack-mcp HYPERSTACK_API_KEY=hs_your_key
npm i hyperstack-core npx hyperstack-core login npx hyperstack-core init openclaw-multiagent
pip install hyperstack-py from hyperstack import HyperStack hs = HyperStack(api_key="hs_your_key")
curl -X POST "https://hyperstack-cloud.vercel.app/api/cards" \ -H "X-API-Key: hs_your_key" \ -d '{"slug":"use-clerk","title":"Use Clerk"}'
# OpenAI embeddings docker run -d -p 3000:3000 \ -e DATABASE_URL=postgresql://... \ -e JWT_SECRET=your-secret \ -e OPENAI_API_KEY=sk-... \ ghcr.io/deeqyaqub1-cmd/hyperstack:latest # Fully local – Ollama (no OpenAI key needed) docker run -d -p 3000:3000 \ -e DATABASE_URL=postgresql://... \ -e JWT_SECRET=your-secret \ -e EMBEDDING_BASE_URL=http://host.docker.internal:11434 \ -e EMBEDDING_MODEL=nomic-embed-text \ ghcr.io/deeqyaqub1-cmd/hyperstack:latest
PRICING

Start free. Scale when ready.

Free
$0
forever
  • 50 cards
  • All features – graph, impact, replay
  • REST API + MCP, 1 workspace
Popular
Pro
$29/mo
billed monthly
  • 500+ cards
  • Semantic search
  • All 4 graph modes
  • GraphRAG search
  • Time-travel
  • 3 workspaces
Team
$59/mo
billed monthly
  • 500 cards
  • Everything in Pro
  • 5 team API keys
  • Unlimited workspaces
Business
$149/mo
billed monthly
  • 2,000 cards
  • Everything in Team
  • 20 team members
  • Agent-to-agent webhooks
  • SSO & audit log (soon)
TRUSTED BY EARLY ADOPTERS

Build agents you can audit and trust.

Provenance is infrastructure,
not an afterthought.

Build agents you can audit and trust at $0/operation. Ledger-like accountability for multi-agent systems.

Join Discord
Free: 50 cards. $29/mo unlocks 500+ cards + branching →