Skip to content

SecondMemory Enterprise

Your company’s memory. Your people’s privacy.

One shared memory for your projects, customers and decisions, ready in every AI tool you approve. The company owns the work. Your people own their lives. Run it in our cloud, in yours, or on your own servers.

Request an invite

Invite only. Request an invite and choose Business, and we’ll be in touch about your team.

Two memories. One clear line.

Work goes to the company’s memory. Life goes to each person’s own. Nobody has to choose what to share.

Owned by your company

Business memory

  • Projects, customers, decisions and deadlines
  • Shared with the people on each project
  • Locked with keys your company holds
  • Stays when people leave

Owned by each person

Private memory

  • Their notes, their family, their health, their life
  • Never shared with the company
  • Locked to their own phone
  • Leaves with them

A private note can point to a work project. Work memory never points back. The company can’t even tell that private memory exists.

You hold the keys.

Not us. Not your AI vendor. You.

Your account. Your keys.

Your company's keys stay in your own cloud account. Not ours.

Delete the key. Delete the memory.

Every byte of company memory becomes unreadable at once. For everyone, including us.

Designed so we can't look.

Memory is only ever opened inside sealed computers that even we can't see into.

Self-hosting

Run it where you want.

In our cloud, in yours, or on your own servers. The same SecondMemory, with the same privacy, wherever it runs.

Managed

Our cloud.

We run it for you in Sydney. Nothing to install or maintain, and your company still holds the keys.

Bring your own cloud

Your cloud.

Deployed into your own cloud account. Your network, your region, your keys. We never get access.

Self-hosted

Your servers.

Run it on your own hardware or private cloud. Pair it with an AI model you host yourself, and your memory never leaves your building.

  • Standard containers
  • Bring your own AI model
  • Your keys, in every option
  • No connection to us required

Admins get control. Not access.

Your admins can

  • Add and remove people
  • Create projects and decide who's on them
  • Choose which AI tools are allowed at work
  • Set retention and legal holds on work memory
  • See who used company memory, and when

Your admins can’t

  • Open anyone's private memory
  • Even see that it exists
  • See which AI tools people use for themselves

People come and go. Knowledge stays.

  1. Joins

    Gets the keys to their projects, in a work profile on their phone.

  2. Moves team

    The old project closes, the new one opens. The old key stops working.

  3. Leaves

    Work memory stays with you. Their private memory goes with them.

  4. You leave us

    Delete your keys, and all of it is gone. Provably.

For engineering teams

Your engineers spend less on AI.

Set your team size and a typical month for one engineer.

Start from
50
40M

Fresh tokens you send that aren't cached

20M

What the model writes back

3B

Context re-read on every turn

200M

Context stored in the cache

Your team's AI bill today

$108,000/ month

You could save up to*

$42,900

a month, or $514,800 a year.

40%off your AI bill*

Input tokens
−$2,400
Output tokens
−$3,000
Cache reads
−$15,000
Cache writes
−$22,500

* An upper-range estimate, not a quote. It assumes SecondMemory lets you work in shorter sessions that carry about half as much context on every turn (cache reads −50%, cache writes −45% after the cost of starting fresh sessions), that you paste in and re-read 30% less (input −30%), and that your agents spend 15% less output looking around for things they already decided. Prices are Anthropic’s public list prices for Claude Opus 5.5 on 27 September 2026. Your savings depend on how you work; the app shows the savings it measures for you.

Questions from IT and security.

Does it work with our single sign-on?
Yes. People sign in through your identity provider, such as Microsoft Entra ID, Okta or Google, and are added and removed automatically when your directory changes. Tell us which one you use when you request an invite.
Where is our data stored?
In our cloud, it’s stored in Sydney. In your cloud, it’s in the account and region you choose. Self-hosted, it stays on your servers. Wherever it’s stored, it’s locked with keys your company holds. The AI model runs wherever that model runs; pair it with a model you host yourself and nothing leaves your network.
Can we bring our own AI model?
Yes, that’s how it’s designed. SecondMemory is the memory, not the AI. Connect the AI tools you already approve, or a model you host yourself.
What does the audit log show?
Who used company memory, through which AI tool, and when. It never shows anything from a person’s private memory, because the company can’t see it.
How do retention and legal holds work?
You set how long work memory is kept, project by project, and you can place a hold so nothing is deleted. Holds apply to work memory only. Private memory belongs to each person.
What happens if we stop using SecondMemory?
Download all of your company’s memory in an open format. Then delete your keys, and everything left with us becomes unreadable, for good.

Give your company a memory.

Business is $299 per seat, per month. Your heaviest AI users could save up to $24,000 a month each on their AI bill.*

Take the next step.

  • Request an invite

    Request one and we’ll be in touch about your team.

    Request an invite

  • See how privacy works

    The keys, the private drawer, and what your admins can’t see.

    Read more

  • Talk to our team

    Pilots, self-hosting, security reviews. Ask us anything.

    hello@secondmemory.io

* Upper-range estimate for the calculator’s “Heavy API user” example on Claude Fable 5 list prices (27 Sep 2026), rounded down; see the assumptions under the calculator above.