Skip to content
💸 usage-based · no seat taxPricing [1/6]

Start free. Pay for what your agents actually use.

No seat tax, no surprise bill. You're metered on three plain things — memories stored, recalls, and how many agents share the memory. Generous free tier; scale only when you do.

Start freeTalk to founder
Plans[2 / 6]

Four tiers. Pick the one your agents are at.

Start free and upgrade only when your usage grows. Pro is where most teams land. Toggle monthly or annual — annual is two months free.

  • Free

    Kick the tires / solo agent.

    $0free forever
    • 5kMemories stored
    • 10kRecalls / mo
    • 1Agents
    • Core remember & recall
    • MCP plugin
    • Community support
    Start free
  • ★ Most picked

    Pro

    A real agent or small team in production.

    $49/ month + usage over the included pool
    • 100kMemories stored
    • 1MRecalls / mo
    • up to NAgents sharing one memory
    • Private + shared buckets
    • Provenance + audit log
    • Higher memory & recall pools
    • Email support
    Start free → upgrade in app

  • Scale

    Many agents, heavy recall, governance that matters.

    $499/ month + volume-priced usage
    • Big poolMemories stored
    • Big poolRecalls / mo
    • ManyAgents
    • Conflict resolution
    • Capability-token identity
    • Data residency choice (EU / US)
    • Priority support
    Start free

  • Enterprise

    Org-wide, governed, with SSO + SLAs.

    Customtailored to your org
    • CustomMemories stored
    • CustomRecalls / mo
    • CustomAgents
    • Everything in Scale
    • SSO / RBAC
    • SLAs + DPA
    • Dedicated support + security review
    TALK TO FOUNDER ↗
What we meter[3 / 6]

Three plain things. No seat tax.

You're metered on memories stored, recalls, and agents — that's it. Here's exactly what each one means.

  • Memories stored

    Facts we keep recallable for you. Compact, deduped — we store distilled facts, not raw transcripts, so this number stays small.

  • Recalls

    Each time an agent asks the memory a question. This is the one that scales with how busy your agents are — and the one we optimize for (sub-200ms p95).

  • Agents

    How many agents (parent + subagents + tools) share one memory. Our wedge — more agents on one governed memory is where the value is, not a penalty.

You see live usage against each meter in the dashboard; nothing auto-charges past your plan without a heads-up.

Compare[4 / 6]

What unlocks where.

The multi-agent and trust rows are where Pro and Scale pull ahead — proof the wedge is what you pay for.

Feature comparison across the Free, Pro, Scale, and Enterprise plans.
FeatureFreeProScaleEnterprise
Limits
Memories stored5k100kBig poolCustom
Recalls / month10k1MBig poolCustom
Agents1up to NManyCustom
Multi-agent
Shared + private buckets—✓✓✓
Conflict resolution——✓✓
Capability-token identity——✓✓
Trust
Provenance—✓✓✓
Audit log—✓✓✓
Poisoning resistance✓✓✓✓
Data residency (EU / US)——✓✓
Integration
MCP plugin✓✓✓✓
SDKs (TypeScript + Python)✓✓✓✓
REST API✓✓✓✓
Proxy—✓✓✓
Support & governance
Support tierCommunityEmailPriorityDedicated
SSO / RBAC———✓
SLA———✓
DPA——✓✓
Questions[5 / 6]

Before you commit.

The questions buyers actually ask about usage, agents, and the bill.

  • What counts as a recall?

    A recall is each time an agent asks the memory a question and gets ranked facts back. It's the meter that tracks how busy your agents are — and the one we optimize for (sub-200ms p95). You see live recall usage in the dashboard.

  • What happens if I go over my plan?

    Nothing auto-charges past your plan without a heads-up. You see live usage against each meter in the dashboard, and we tell you before anything changes — so there's no surprise bill.

  • Do subagents count as separate agents?

    No. Subagents share the parent's memory and identity, so they don't bill as separate agents. More agents on one governed memory is the value, not a tax — the agent meter counts the agents that share a memory, with subagents inheriting the parent.

  • Can I bring my own model? Does it change price?

    Yes, you can bring any model — OpenAI, Claude, Gemini, open models. ULTRAMEMORY is model-agnostic, and your choice of model does not change the price. You're metered on memories, recalls, and agents — not on the model.

  • Is there an annual discount?

    Yes. Pay annually and get two months free versus paying monthly. The toggle above the plans switches every card between monthly and annual pricing.

  • Free → Pro, do I lose data?

    No. Upgrading from Free to Pro keeps your existing memories — you simply unlock higher pools, shared buckets, provenance, and audit. You can upgrade in the app the day your agents outgrow Free.

  • How do you store so little — what's a “memory”?

    A memory is a distilled fact, not a raw transcript. We dedupe and compact what your agents learn into recallable facts, so the stored count stays small — which keeps your bill small, too.

  • EU / US data residency — which plans?

    Data residency choice (EU or US) is available on Scale and Enterprise. If residency is a requirement for your org, Scale is the entry point, and Enterprise adds SSO, SLAs, and a DPA.

Still deciding? See how we keep memory safe & trustworthy, read how recall stays fast or why your model is your choice, or browse the developer docs.

Pricing [6/6]

Start free. Upgrade the day your agents outgrow it.

Free to start, usage-based, no seat tax. Pay only when your agents and recalls grow.

Start freeTALK TO FOUNDER ↗