Plain answers.

SMTRY is a San Francisco software company that builds AI infrastructure people own outright. Its first product, Anamnesis, is a memory for the AI tools you already use. SMTRY Applied runs fixed-scope engagements on that stack, and SMTRY Labs is where the research happens. SMTRY is a venture of J. I. Ashley Consulting.

A persistent, encrypted memory for AI assistants. It keeps what you tell Claude, ChatGPT, Codex or Gemini across sessions and across products, so the context you built in one tool on Tuesday is there in another on Thursday without re-explaining it. It is live in free alpha: it runs, people use it, and it is still changing.

Seven clients through one MCP connector: Claude Code, Claude Desktop, claude.ai, Cowork, ChatGPT, Codex CLI, and Gemini CLI. One account, one memory, shared across all of them.

One honest distinction: fully automatic session capture runs in Claude Code today. The other tools read your memory into every chat and save new memories when you ask ("remember this") or when the assistant judges something worth keeping.

What you say to your assistant is recorded as a session, cut into episodes about one thing each, distilled into short typed memories (a decision, a preference, a reference), and consolidated over time into long-term knowledge. When you start a new conversation, its context is the cue: the memories that bear on it are placed in front of the model before it answers. Warm recall takes 0.2 to 0.4 seconds, and a distilled memory carries up to 27 times its stored size in source context. The stages and their sources are on the architecture page; the numbers link to the benchmarks.

Nothing during the alpha, with no card required. The published tiers: Free (a hosted trial, and local self-hosting free forever), Hero at $19 per month for a solo developer, Demigod per seat for teams (shipping soon), and Enterprise by conversation. There is no ad tier and no data-sale tier, at any price. Prices are pre-launch and may change before paid tiers ship; details on the pricing page.

Each account's memory is encrypted with a key derived from its own credentials. There is no master key, so a database-only breach decrypts nothing, and a leaked credential exposes only that account. SMTRY does not train models on your memory, does not sell it, and runs no third-party trackers.

What we do not claim: Anamnesis is not end-to-end encrypted today. The server holds your API key to authenticate you and could derive your key from it. Client-side key derivation is on the roadmap. The full threat model, with what it does and does not defend against, is on the security page.

Yes. The dashboard lists every memory, searchable, and any one can be deleted. Export downloads your whole corpus as plain JSON with no proprietary format in the way. Delete your account and the encryption key is destroyed, so even rolling backups become unreadable.

Yes. The free tier includes local mode: self-host the MCP server on your own hardware. For organizations, SMTRY Applied installs the memory stack inside your network with keys held by you, and the Enterprise tier offers hosted, hybrid, or fully on-premises deployment. In an August 2026 holdout, an 8-billion-parameter open-weight model on one consumer machine matched the cloud pipeline's retrieval quality, 0.94 evidence coverage on both arms; the trade today is time. Labs has the measurement.

SMTRY Applied is the engagements division: it puts the memory stack into the systems you already run. Two engagements are listed: memory integration, priced as a fixed fee scaled to the seats involved so it reads as a product purchase with setup, and sovereign deployment, priced as a fixed fee scaled to the hardware it runs on plus a support retainer. Custom builds that extend the stack are scoped by conversation and never listed.

Staff augmentation. Hourly work. Generic applications. Anything not on SMTRY's own stack, unless it is a project selected for a reason we can name. Strategy, brand and marketing, web design and operations, and AI search visibility are J. I. Ashley Consulting's work, and we send them there by name.

J. I. Ashley Consulting founded SMTRY and is its sibling firm for business-focused advisory: strategy including AI strategy, program delivery, brand and marketing, web design and operations, and AI search visibility. SMTRY builds and runs the software. Neither firm builds generic custom software, and when a client's work belongs with the other, each says so and hands it across by name.

Not yet. Behind Anamnesis runs a SOC 2-aligned readiness program: a control matrix mapped to the Trust Services Criteria, written policies, and weekly evidence snapshots committed to a tamper-evident history since August 2026. No certification is claimed until an independent CPA firm signs one. The audit starts when it is funded.

Atria is the second product and is in development; we do not describe what it does until there is a measurement to describe. Craft is an alpha we run for our own writing desk rather than sell: an editing partner that learns a writer's instincts and pushes back where habit is standing in for a choice. It is not available outside the firm.

Tell your connected assistant to "report an issue with Anamnesis" and it files a structured report. Or write to hello@smtry.ai, call (415) 644-8604, or use the contact page. Security concerns go to security@smtry.ai. A real person reads every report.