hyperstack-py 1.5.4

The Agent Provenance Graph for AI agents. Timestamped facts, auditable decisions, deterministic trust. Prove what agents knew, trace why they knew it, coordinate without an LLM in the loop. $0 per operation.

🃏 HyperStack Python SDK

The Agent Provenance Graph for AI agents — the only memory layer where agents can prove what they knew, trace why they knew it, and coordinate without an LLM in the loop. $0 per operation at any scale.

Timestamped facts. Auditable decisions. Deterministic trust. Build agents you can trust at $0/operation.

Install

pip install hyperstack-py

Current version: v1.5.3

Quick Start

from hyperstack import HyperStack

hs = HyperStack("hs_your_key")

# Store a memory
hs.store("project-api", "API", "FastAPI 3.12 on AWS", stack="projects", keywords=["fastapi", "python"])

# Search memories
results = hs.search("python")

# Get a single card
card = hs.get("project-api")

# List all cards
cards = hs.list()

# Delete a card
hs.delete("project-api")

# Graph traversal (what blocks X? what does X affect?)
blockers = hs.blockers("deploy-prod")
impact = hs.impact("use-clerk")

# Time-travel: graph at a past timestamp
graph = hs.graph("auth-api", at="2026-02-01T00:00:00Z")

# Utility-weighted edges (report success/failure)
hs.feedback(card_slugs=["use-clerk"], outcome="success")

# Git-style branching
branch = hs.fork(branch_name="experiment")
hs.diff(branch_workspace_id=branch["branchWorkspaceId"])
hs.merge(branch_workspace_id=branch["branchWorkspaceId"], strategy="branch-wins")

# Agent identity + trust
hs.identify(agent_slug="research-agent")
profile = hs.profile(agent_slug="research-agent")

# Ingest conversation transcript into cards
hs.auto_remember("Alice is a senior engineer. We decided to use FastAPI over Django.")

# Memory hub: working (TTL) / semantic (permanent) / episodic (30-day decay)
cards = hs.hs_memory(surface="semantic")

# Batch store multiple cards
hs.bulk_store([{"slug": "p1", "title": "Project A", "body": "..."}, {"slug": "p2", "title": "Project B", "body": "..."}])

# Parse markdown/logs into cards (CLI + programmatic)
hs.parse("# DECISIONS.md content or log output", source="decisions")

# Agentic routing: deterministic, no LLM
can_do = hs.can("auth-api", action="deploy")    # Can this card do X?
steps = hs.plan("auth-api", goal="add 2FA")     # Plan steps for goal

API Reference

Method	Description
store(slug, title, body, ...)	Create/update a card
bulk_store(cards)	Batch store multiple cards
get(slug)	Get one card
search(query)	Search cards
list(stack=None)	List all cards
delete(slug)	Delete a card
graph(from_slug, depth, at, reverse)	Forward/reverse traversal + time-travel
blockers(slug)	What blocks a card
impact(slug)	Blast radius of a change
feedback(card_slugs, outcome)	Report success/failure, updates utility scores on edges
fork(branch_name)	Git-style branch
diff(branch_workspace_id)	Compare branch to parent
merge(branch_workspace_id, strategy)	Merge branch
discard(branch_workspace_id)	Delete branch
identify(agent_slug)	Register agent identity
profile(agent_slug)	Get agent trust score
can(slug, action)	Agentic routing: can this card do X? (deterministic, no LLM)
plan(slug, goal)	Agentic routing: plan steps for goal
auto_remember(text)	Ingest conversation transcript into cards
hs_memory(surface)	Memory hub: working / semantic / episodic
parse(text, source)	Parse markdown/logs into cards

Card Fields

Field	Description
confidence	0.0–1.0 confidence score
truthStratum	draft | hypothesis | confirmed
verifiedBy	e.g. "human:deeq"
verifiedAt	Auto-set server-side
memoryType	working | semantic | episodic
ttl	Working memory expiry (seconds)
sourceAgent	Auto-stamped after identify()

Backend Features

• Conflict detection — structural, no LLM, auto-detects contradicting cards
• Staleness cascade — upstream changes mark dependents stale
• Three memory surfaces — working (TTL), semantic (permanent), episodic (30-day decay)
• Decision replay — reconstruct agent state at decision time + hindsight detection
• Time-travel — graph() with at= timestamp
• Self-hosting — Docker + HYPERSTACK_BASE_URL env var

Why HyperStack?

• Provenance tracking — timestamped facts, auditable decisions
• Decision replay — reconstruct what agents knew at decision time
• Deterministic trust — no LLM in the loop for coordination
• $0 per operation — at any scale
• Zero dependencies — just Python stdlib
• 30-second setup — get key at cascadeai.dev/hyperstack

Pricing

Plan	Cards	Price
Free	50	$0/mo — ALL features including graph
Pro	500+	$29/mo
Team	500, 5 API keys	$59/mo
Business	2,000, 20 members	$149/mo

Get a free key: cascadeai.dev/hyperstack

License

MIT © CascadeAI