Free
For early experiments and tiny private tools.
$0100 active records
- 100 active records
- Dashboard and API access
- Archive records anytime
Portable context for every AI. Never explain yourself twice.
A shared memory layer for your AI tools. Tell one something, and they all know.
Connects to the tools you already use
ClaudeChatGPTCodexOpenClawHermesManusYour custom app
The problem
You're the integration layer. That's the problem.
You explain your project to one AI. Then you open another and explain it all over again. Claude, ChatGPT, Codex — none of them share a memory, so you're stuck copy-pasting your own context between tabs.
Cpaste again
Claude
"We are a B2B startup building a budgeting app for small businesses…"
Gpaste again
ChatGPT
"We are a B2B startup building a budgeting app for small businesses…"
Cxpaste again
Codex
"We are a B2B startup building a budgeting app for small businesses…"
AO2Memory
ChatGPT
Codex
OpenClawYour appsvia API
ClaudeThe solution
One memory. Every tool.
AO2 Memory is a shared memory layer for your AI tools. It's an MCP server you connect to whatever you use — Claude, ChatGPT, OpenClaw, your own apps via API — so they all read and write to the same place. Tell one agent something, and the rest already know.
Everything you'd want from memory you actually own.
Mention it once. Every agent knows.
Connect via MCP, API key, SDK, or CLI.
Works with Claude, ChatGPT, Codex, OpenClaw, Hermes, Manus — even your own apps.
Claude
ChatGPT
Codex
OpenClaw
Hermes
Manus
Your custom app
Connect however you build
MCPAPI keySDKCLI
Activity
- UpdatedLisbon tripChatGPTJun 24Roll back to v2
- CreatedStandup notesClaudeJun 23
- ChangedTasks schemaCodexJun 22
See what every agent did. Roll back anything.
A dashboard of every change an agent made to your memory.
Undo any of it, anytime. No more wondering what an AI touched.
Control exactly what each AI can touch.
Scope access down to the record and field.
Hand an agent only what it needs. Issue temporary keys that expire on their own.
ChatGPTaccess keyActive- titleVisible
- statusVisible
- salaryHidden
- ssnHidden
Allowed actions
records:readrecords:updaterecords:archivestructure:write
Temporary key · expires in 7 days
Free-form blob
lisbonbudget?todosarahjune??statusidkflightnotespriorityhotelmaybe…
AO2 — structured records
titletext
statusselect
duedate
prioritynumber
ExportYours
Structured memory, not a messy text dump.
Define your structure once and every record follows it.
Your notes stay legible to AI instead of rotting into an unsearchable pile.
Your memory walks out the door with you.
Export everything with one command.
Push it to Git, or pull it down as JSON, Markdown, or CSV. No lock-in, no platform holding your context hostage — it's yours to keep.
Export memory
$ ao2 memory export --gitdone
pushed 142 records to your repo
GitJSONMarkdownCSV
Captured automaticallyAuto
- Prefers window seats on flights
- Ships on Thursdays
- Allergic to penicillin
248memories and growingcompounding
It remembers, so you don't have to.
Agents save what they learn as you work.
Your memory compounds over time. No manual saving, no copy-paste.
How it works
Set up in under a minute.
1
Connect
Plug AO2 into any AI tool via MCP, or grab an API key.
2
Seed it
One prompt and your memory populates itself — or start from a template.
You → “Analyze what you know about me and load it into AO2.”
3
Just work
From here it compounds — agents save what they pick up, and every tool you connect already knows.
Pricing
Start small. Upgrade when your records become real work.
Plans are based on active records across your account. Archived records stay available without counting toward the limit.
Pro
For active apps with room to grow.
$15/mo1,000 active records
- 1,000 active records
- Stripe-managed billing
- Room for active apps
Unlimited
For serious AO2 workspaces with no record ceiling.
$30/moUnlimited active records
- Unlimited active records
- Best for serious workflows
- No record ceiling
Stop explaining yourself to every AI.
Set up your shared memory layer in under a minute.
Brought to you by AO2
The deployment stack for “audience-of-two” apps.
A new kind of software — built for just two: you and your agent.
Personal software changed. AI can build almost anything now — a tool, an app, a workflow, in minutes. But the moment you try to actually use what you built, you hit the same wall every time.
It's not reachable outside your laptop. You need auth so the whole internet can't poke at it. You need a domain. You need a database you can trust and audit. All of that is hours of plumbing for something you just wanted to spin up and use.
That stack doesn't exist yet for small, personal, two-user apps — software meant for an audience of two: you and your agent. So we're building it.
AO2 Memory is the first piece. The rest is coming.
Want to follow what we're building? Leave your email.
No spam. Just the occasional update.