Use Cases for Asset Operations and AI Cognitive Memory
Version control for creative work. Persistent, auditable memory for AI agents. Both built on the same graph.
For Creators
Keep every prompt, model, and revision connected—while files stay on your machine or NAS.
- ✓ Find any past output in seconds by context, not folders
- ✓ Preserve prompts, seeds, and provenance automatically
- ✓ Publish confidently with revision history and lineage
For Studios
Use Kumiho as a history backbone for production lineage, approvals, and automation - without moving files.
- ✓ Track lineage across DCCs and pipelines with typed edges
- ✓ Immutable revision ledger for reproducibility and audits
- ✓ Events to trigger renders, reviews, and publishing workflows
AI Cognitive Memory
Give your AI agents persistent, auditable memory that survives context resets, model swaps, and long-running projects.
- ✓ Graph-native memory with typed reasoning edges
- ✓ Dream State consolidation for overnight enrichment
- ✓ MCP plugin for Claude, Cowork, and agent frameworks
from kumiho import memory_store, memory_recall
# Store a decision with provenance
memory_store(
summary="Approved hero layout",
edges=[{"type": "DERIVED_FROM",
"target": "kref://project/brief/rev/2"}]
)
# Recall by meaning
results = memory_recall("hero layout decision")
print(results[0].summary, results[0].edges)For Developers
Build automation and tooling on a clean API: revisions, edges, search, events, and MCP tools.
- ✓ SDK-first interfaces (Python/C++/Dart) for tooling
- ✓ Streaming events for orchestration and notifications
- ✓ MCP endpoint for agent and workflow integrations
from kumiho import Kumiho
client = Kumiho(url="http://localhost:8000", token="...")
# Create a revision and connect lineage
rev = client.revisions.create(item_kref="kref://.../my-item", tag="latest")
client.edges.create(source=rev.kref, target="kref://.../source", edge_type="DERIVED_FROM")
# Stream events for automation
for event in client.events.stream(kref_filter="kref://project/space/**"):
print(event.routing_key, event.kref)Common primitives across all use cases
Kumiho stays consistent no matter your scale: immutable revisions, typed edges, external artifacts, and events.
Immutable revisions
Every change becomes a revision you can tag, audit, and reference.
Lineage edges
Connect revisions with typed edges like DERIVED_FROM or DEPENDS_ON.
Bring-your-own storage
Track artifacts by reference; keep files on disk, NAS, or object storage.
Events for automation
Stream changes to drive render farms, publishing, and notifications.
Workflow stories
A few concrete patterns teams use Kumiho for today.
Creators: prompt-to-publish provenance
Generate in ComfyUI, track prompts and outputs as revisions, then publish a clean “approved” tag.
Studios: lineage across departments
Connect plates → comps → finals with typed edges, keep artifacts in your existing storage, and audit approvals.
AI agents: persistent recall across sessions
Store conversation summaries, retrieve context by query, and let Dream State prune and enrich memory overnight.
Developers: event-driven automation
Stream events into n8n for notifications, publishing, or downstream compute without polling.
Proof and trust
See how Kumiho fits into local-first workflows with a revision ledger, lineage, and automation signals.
Ready to see Kumiho on your pipeline?
Tell us what you are building—creator workflows, studio pipelines, AI agent memory, or developer tooling—and we will help you map Kumiho into it.