Why we built Patina
You're the name
Ask any team going all-in on agents who's actually figured them out. There's usually one or two of you in the team, and you've managed to get great agent output. Your agents have the context they need to sound like your company and know who it's for.
With that challenge done, you did the responsible thing. You wrote it down to share it with the rest of the team. One shared brand voice document, the buyer personas and positioning, the working prompts, the rules for what good looks like. You walked the team through it. However, within weeks there are multiple versions of a single SOP, everyone still has different prompts, and the corrections still come to you.
What sharing is missing
For your team to work from something, four things have to be true:
- They can tell which version is current.
- Someone notices when two entries stop agreeing.
- Every change carries a record of who said yes to it.
- And people can contribute without breaking what already works.
A plain doc gives you almost none of that. A wiki gets you partway (a current page and a history), but when it comes to noticing when two pages stop agreeing, or recording who vouched for a change, few tools do either.
Now introduce agents, and the stakes rise. An agent treats every version as equally true and amplifies whichever one it's handed. That's why it's so hard to keep agents consistent: an agent follows a superseded SOP without noticing. Trust isn't a property of what's written. It's a property of how changes happen.
Owned by the team
So we built Patina for this problem. Patina is one place your agents and your teammates can collaborate. Teammates propose changes in plain language. A change lands only when a reviewer says yes; teammates can't approve their own change, and every approval stays on the record. All without needing to know developer tools like git.
Patina goes a step further: the knowledge is actively maintained. Agents run nightly to scan and flag entries that may conflict or have gone out of date.
You don't need to change your workflow at all. Your agents read from the library live, through the same tools they already call, so nobody is pasting context into a chat window again. The library belongs to the company and stays portable: your agent can export every entry back out as plain markdown.
You'll no longer be the blocker to contributions; they wait for your yes, and a yes takes minutes. You keep the part you'd keep anyway: deciding what's true.
We run on it
We run on Patina ourselves. Most of our contributors are agents. They draft files in a working area, and they only get promoted to the trusted library when a person approves them. The library behind this page holds 367 entries this week. Our voice guide is on its sixth approved version; each one replaced the last through the same review queue. The nightly scan flagged pairs of entries that had quietly stopped agreeing. Two of those disputes are resolved on the record, and one is still open.
One thing isn't built yet: resolving a dispute lives in the web UI today.
The founding 20
We're taking 20 founding teams and setting each one up personally: your workspace, your rules, and we walk your existing context in with you, whether it lives in a repo, a wiki, or that doc, contradictions included. The import tooling is built; the judgment calls aren't. Deciding what's current, what's superseded, and where two documents quietly disagree is work we do together, so we cap the cohort at what we can do well. That's the whole reason for the number.
It's $149 a month for the workspace, everyone included, fourteen days free. When you click, you get a workspace, and we schedule your setup session. What you built carries over; this is where it graduates. The system becomes the company's, with you as its reviewer. What you figured out becomes what the company knows.