Pricing built for asset operations - now with Cognitive Memory for AI agents.

Kumiho is a graph-native platform for revisions, lineage, and automation - plus optional AI Cognitive Memory with long-term recall, provenance, rollback, and audit trails.

Works with your storage and workflows. Built for traceability, not just chat history.

New

30 days of Studio, free.

Your 30-day Studio trial activates the moment you save your first memory — 500,000 nodes, cross-session recall, audit visibility. No credit card. At the end, choose a cloud plan or move your work to Community Edition.

Refer Kumiho — earn 30 more days per signup, stackable up to 90.

Two products, one system.

Kumiho Platform manages assets, revisions, lineage, and automation.

AI Cognitive Memory turns the same graph primitives into long-term memory for agents (episodic to semantic consolidation, dependency-aware updates, auditable reasoning).

LLM model layer is decoupled

Kumiho does not provide or resell LLM inference. Connect your own OpenAI, Anthropic, local model, or AI agent runtime. Kumiho provides memory, provenance, indexing, and auditability.

How buying works

  • AI Cognitive Memory is included with every hosted plan.
  • Bring your own LLM or AI agent stack; Kumiho is the memory layer.
  • Studio trial activates on first memory capture — 30 days, no credit card required.

Kumiho Platform plans

Core platform tiers for creators, studios, and teams.

Creator

$40/month

For solo creators and freelancers

Manage personal workflows, versions, and outputs with full memory capabilities.

  • Core graph workspace (assets + revisions)
  • 30,000 nodes
  • AI Cognitive Memory included
  • Higher limits for revisions, tags, and organization
  • Priority performance tier

Studio

Most Popular

$99/month

For small teams shipping real work

Shared knowledge, traceable decisions, and agent workflows that do not forget.

  • Everything in Creator
  • 500,000 nodes
  • Team collaboration and permissions
  • Cross-session recall and time-travel memory queries
  • Audit visibility for recall reasoning

Studio Pro

$170/month

For high-throughput studios and production environments

Heavier automation, larger memory graphs, and faster retrieval on high-performance infrastructure.

  • Everything in Studio
  • 2,000,000 nodes
  • High-performance graph infrastructure
  • Advanced memory controls
  • Priority support and SLA-ready operations

Enterprise

Contact

For org-wide deployment and compliance

Security, governance, and custom architecture at enterprise scale.

  • Custom limits, compliance, and security controls
  • Dedicated onboarding and architecture support
  • Multi-team governance and advanced reporting

All platform plans include BYO storage. You keep control of your files.

Free · Self-hosted

Prefer to self-host? Run Community Edition.

Community Edition replaces the retired cloud Free tier. It is a free, single-user build you install on your own machine, with no account, token, or cloud connection.

macOS / Linux — one command:

curl -fsSL https://github.com/KumihoIO/kumiho-server-community/releases/latest/download/install.sh | sh

Community Edition limits

  • Single-user, loopback-only (binds to 127.0.0.1)
  • No data caps — your local graph is unlimited
  • Requires a local Neo4j 5.x; Redis 7.x optional
  • Not for hosted, team, or production-backend use

What is included in AI Cognitive Memory?

Step 1

Working Memory

Session-local context in Redis for fast store/retrieve operations.

Step 2

Long-term Graph Memory

Consolidated facts and links persisted with provenance edges.

Step 3

Dream State

Scheduled consolidation refines memory and supersedes stale beliefs.

See detailed limits

Memory Ops, retention, cadence, and graph scale.

LimitCreatorStudioStudio ProEnterprise
Nodes (total)30k500k2MUnlimited
AI Cognitive MemoryIncludedIncludedIncludedIncluded (Custom)
Working memory retention (per session)24h24h72hCustom
Memory ingest / day500IncludedIncluded (higher)Custom
Max buffered messages / session2001,0002,000Custom
Dream State (nightly consolidation)NightlyNightly2x nightly or deeper passCustom cadence
Memory retrieval modeHybrid (BM25 + vector)Hybrid (BM25 + vector)Hybrid (BM25 + vector)Hybrid (BM25 + vector)
Vector-indexed memory items / month10k100k250kCustom
AI Memory Ops / month25k150k400kCustom
API rate limit (req/sec)100100100Custom
Events / day2,0005,00010,00050,000

AI Cognitive Memory uses the same graph primitives and counts toward your node limit.

BYO-LLM: Kumiho charges for infrastructure and indexing. LLM usage runs on your own provider key.

AI Memory Ops / month: Memory Ops = store + retrieve + consolidate. These can map to event stream throughput.

FAQ

Security, privacy, and billing behavior at a glance.

What makes AI Cognitive Memory different from RAG or vector memory?

It stores structured revisions and explicit relationships so agents can trace sources, revise beliefs, and explain recall decisions.

Where is the free option?

The retired cloud Free tier has been replaced by the self-hosted Community Edition.

What does belief revision mean?

New information can supersede older memory while preserving linked history for traceability and rollback.

What happens if I exceed limits?

Add usage packs, upgrade tiers, or apply throttles and retention policies.

What happens to my memories if I stop using Kumiho?

Inactive accounts move to cold storage after 90 days — we never silently delete your memories. Log back in at any time and the graph re-indexes within minutes.

Can I earn more Studio time?

Yes — refer Kumiho and earn 30 days of Studio per successful signup, stackable up to 3 (90 days max). Beyond the cap, additional referrals convert to account credit.

Memory at every tier. Pay for scale.

Start with Creator or Studio in the cloud, or run Community Edition on your own machine for free.

Kumiho Pricing — AI Agent Memory Plans & Community Edition