Beta
Your context, enforced on everything your AI writes.
Confirm your facts, rules and skills once, and they follow you into Claude, ChatGPT and whatever you use next. Datum checks every draft against them and holds back what doesn’t match before it goes out in your name.
Free to start. No card needed.
Your prices. Your dates. What you confirmed.
Your brand guide, your call notes. What you work from.
Words you never use. Limits you never cross.
One space per project. Nothing crosses over.
The steps you take every time. Saved once.
A signed record of every check. It can’t be quietly changed.
One account, the same facts and rules in every AI you use
- Claude
- ChatGPT
- Grok
Gemini- Perplexity
Manus
Claude, ChatGPT, Grok and Manus let you add a custom MCP connector on their free plans. Perplexity requires a paid plan, and Gemini support is coming.
What people are saying
What users say, post and write about it, in their own words.
Saying it
“Datum skills automate my posting, so I’m not writing every post by hand. Datum checked nearly every one of the 400+ posts I’ve scheduled before it went out. Now I’m getting calls from people who saw them.”
“Even my system processes, workflow and everything has always just felt heavy when I think about it… That weight is lifted right now. Even though I got work to do still, it feels more like manageable and doable.”
“It’s definitely helping, because it already knows. It makes using Claude at least five times faster.”
“This is the need of the market right now. It’s useful for enterprises, it’s useful for agencies when they implement agents, it’s useful for everybody.”
“Datum has done a great job of keeping my facts and document organization consistent and well structured while aligning with my personal style.”
The difference
Everything else informs your AI. This one refuses.
A prompt asks and hopes. A memory tool reminds. Neither one stops a wrong number going out, so you end up checking anyway. Datum measures the finished draft with code, not another AI, so the same draft gets the same answer every time. It refuses what doesn’t hold and leaves a receipt anyone can check.
How it works
Point it at what you’ve written. Confirm what’s right. Every draft gets checked.
Datum proposes your facts, rules and skills from what you already wrote. Nothing is used until you confirm it, and from then on every draft is checked against it.
One correction, one approval. Nothing is used until you say so.
Put it to work
Your skills do the work. Datum checks every word before it goes out.
A skill is the steps you’d follow, written once. These are the two I run on my own LinkedIn. My AI does the work, and nothing posts until Datum clears it.
Every day
Comment where my buyers are talking
ClaudeNimbleDatumLinkedIn
- 1Find people outside my network posting about a problem Datum solvesNimble
- 2Read the postNimble
- 3Draft one sentence with my opinion, from facts I confirmedClaude
- 4Datum checks it. Twelve words or fewer, none of my banned words, nothing I haven’t confirmedDatum
- 5Post it. When someone replies, I take overLinkedIn
Every week
A week of posts, written and scheduled
ClaudeDatumMetricool
- 1Pick the best fourteen from the post ideas I’ve savedDatum
- 2Write each one the way my LinkedIn post skill saysClaude
- 3Datum checks every draft against my rules and facts, and holds back any that don’t matchDatum
- 4Schedule them twice a day for the weekMetricool
Both run on a schedule in Claude. Nimble finds and reads the posts, Metricool schedules mine, and Datum holds the skills, facts and rules and checks every word. My comments post through a LinkedIn connection inside Datum that only I have for now. Until it’s released, your AI can post them through your own LinkedIn tools.
Templates
Start from how other businesses already use it.
Paste one to your AI. It asks you the questions, and the rules come out yours. No one else’s numbers ship inside them.
Why this exists
When your AI gets it wrong, it is your problem.
-
78%
of small business owners don’t trust AI to perform basic tasks. 74% are using or testing it anyway.
Bluevine 2026, 942 US small business owners -
40%
of the time AI saves you goes back into fixing what it wrote. For a daily user, a week and a half a year.
Workday 2026, 3,200 employees -
Liable
A tribunal held Air Canada responsible for a policy its chatbot invented. It argued the bot was separate. That failed.
Moffatt v. Air Canada, 2024
Datum doesn’t stop your AI being wrong, and it’s no replacement for reading your own work. It makes what your AI wrote checkable by someone who wasn’t there. What goes out still goes out in your name.
See it
What it caught for beta users.
Real catches from the first weeks of the beta, with names and details left out. Each one looked fine until something checked it.
The description gave the lake frontage as about four times what the MLS showed.
Datum flagged itChecked against the MLS on the same page. The agent had no idea the two disagreed.
“Transparent pricing,” on a website with no prices anywhere on it.
Datum flagged itOne of seven places the company’s own website contradicted itself.
An old keynote price, still on the site.
Datum flagged itIt no longer matched the price he’d confirmed in Datum. Caught while drafting an update to the site.
Pricing
Start with your facts. Add skills. Bring your people in.
Start free on one space. On your own or with staff, it’s the same Datum, and each step up adds one thing.
Try Datum on one area of your work.
Set it up- ✓1 space, plus 1 subspace inside it
- ✓3 rules
- ✓1 source
- ✓1 skill
- ✓Unlimited material
- –No shared access
- ✓50 MB storage
- ✓Unlimited checks, receipts and export
- –Content may be used to improve Datum
Your rules and context, unlimited.
Choose Plus- ✓Unlimited spaces, rules and sources
- ✓Unlimited material
- –No skills
- ✓5 GB storage
- ✓Unlimited checks, receipts and export
Plus, with skills to build automations you can trust.
Choose Pro- ✓Everything in Plus
- ✓2 people, working to the same confirmed rules
- ✓Unlimited skills
- ✓Vault, proof of what each check was based on
- ✓Every confirmation names who made it
- ✓20 GB storage
- ✓Unlimited checks, receipts and export
- ✓Everything in Pro
- ✓5 people, working to the same confirmed rules
- ✓$75 per extra person
- ✓Approval chainsSoon
- ✓Roles and an adminSoon
- ✓Team audit viewSoon
- ✓50 GB storage
- ✓Unlimited checks, receipts and export
Already use Datum? Sign in to upgrade or manage your subscription. Compare every plan, or if you set Datum up for clients, see agency pricing.
Billed monthly by card. Cancel any time. Anyone who joined before October 5, 2026 keeps the plan and price that was listed when they joined for a year, with thirty days notice before anything changes.
Before it goes out in your name,
make sure it holds up.
Free to start, no card needed. I’m onboarding every new user one on one.
The free plan needs no card.
Questions
The ones people actually ask.
Does this replace Claude or ChatGPT?
No. It runs inside whichever one you already pay for. Datum performs no thinking of its own and has no model. Your assistant writes, and Datum decides whether what it wrote matches what you confirmed.
Do I need to be technical?
No. If you can use ChatGPT or Claude, you can use Datum. Setup takes one connection and a few settings, and I’ll do it with you on a call if you’d rather not do it alone.
How is this different from giving my AI a better memory?
A memory tool feeds context in and hopes. None of them refuse anything. Datum measures the draft against what you confirmed and blocks it when they disagree, which is a different job and the reason a receipt is worth anything.
Does Datum know whether my facts are true?
No, and it never claims to. It knows what you confirmed. If you confirm the wrong price, it will hold your writing to the wrong price faithfully. What it catches is your AI writing something that doesn’t match what you confirmed.
Who can read what I put in?
During the beta everyone is on the free plan, and what you put in may be read by me to make Datum better. It’s never sold and never trains anyone else’s model. When paid plans arrive, they’ll come with a written commitment that nobody reads your content. Either way it’s stored under your account, and you can export or delete all of it at any time. The privacy policy has the details.
Where is my data stored?
On one machine in Fargo, North Dakota, reached over an encrypted tunnel, and backed up daily on the same premises. Sign in is handled by WorkOS, so Datum never sees your password. That’s a beta setup, so don’t put regulated or highly sensitive data in it yet. The privacy policy has the details.
What is a receipt, and why would anyone care?
A signed record of what was checked, against what, and how it came out. If anyone changed it afterwards, including us, the signature would no longer match.
What happens when it’s wrong?
Two ways. Sometimes it refuses a draft that was fine, usually because a rule was written too tightly, and you loosen the rule. And it only checks what your AI sends it, so a claim that never goes through a check isn’t covered.
Is it really free?
Yes. Free covers one space and one skill, with checks and receipts never capped, and no time limit. Plus adds unlimited spaces and rules, and Pro adds unlimited skills. Paid plans bill monthly and can be cancelled any time, and anyone who joined before October 5 keeps the price they joined at for a year. See the plans.