> ## Documentation Index
> Fetch the complete documentation index at: https://docs.memorycrystal.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Pricing

> How Memory Crystal is packaged today, how plan limits show up in the repo, and what changes across tiers.

Memory Crystal is packaged as hosted plans for managed cloud usage, plus local or self-hosted deployment paths for operators who want to run their own backend.
Pricing controls capacity for stored memories, short-term messages, channels, raw sensory retention, asset storage, and managed Gemini usage.

## Current hosted plans

| Plan       |   Price | Core limits                                                                                                                                                   |
| ---------- | ------: | ------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Free       |  \$0/mo | 500 memories, 500 short-term messages, 3 channels, 7-day message retention, 7-day raw sensory retention, 250 MB multimodal asset storage                      |
| Pro        | \$29/mo | 25,000 memories, 25,000 short-term messages, unlimited channels, 30-day message retention, 14-day raw sensory retention, 2 GB multimodal asset storage        |
| Ultra      | \$79/mo | Unlimited memories, unlimited short-term messages, unlimited channels, 90-day message retention, 30-day raw sensory retention, 10 GB multimodal asset storage |
| Enterprise |  Custom | Custom limits, custom retention, custom multimodal asset storage, multi-seat/team usage, custom deployment options, dedicated support and SLAs                |

Pro and Ultra use 7-day trial checkout links on the public pricing page. Enterprise is contact-only.

## How it actually works

The public website renders current plan limits from the shared tier model where possible. Treat `shared/tierLimits.ts` as the repo-level source of truth for memories, short-term message caps, channel caps, retention windows, and managed Gemini behavior.

Important source files:

* `shared/tierLimits.ts`
* `apps/web/app/pricing/page.tsx`
* `docs/PRICING.md`
* `README.md`
* `convex/crystal/geminiGuardrail.ts`
* relevant backend enforcement paths in `convex/crystal/*`

## Plan behavior notes

* `starter` exists in `shared/tierLimits.ts` as a legacy Pro-style alias. It is not a separate public hosted plan.
* Free does not include managed Gemini. Pro includes managed Gemini with a 500-call daily cap. Ultra and Enterprise-style unlimited tiers allow BYOK/custom caps and have no default managed Gemini daily cap.
* Raw uploaded files are not exposed directly to model context today. Recall uses derived text, transcripts, summaries, and metadata.
* Local or self-hosted setup is a deployment mode. Do not assume hosted-tier behavior, support, backups, migrations, or tunnel behavior unless the current install docs and code support it.

## Common mistakes

* treating marketing copy as the technical source of truth for limits
* listing `starter` as a current public plan
* assuming hosted-tier behavior and self-hosted behavior are the same thing
* forgetting that memory limits, message retention, raw sensory retention, asset storage, and model usage are separate controls

## Source of truth

Primary files behind this page:

* `shared/tierLimits.ts`
* `apps/web/app/pricing/page.tsx`
* `docs/PRICING.md`
* `README.md`
* `apps/web/app/data/releases.ts`