SOP & Knowledge Base
Someone walks through a task once — a recorded screen share, a call, a long answer typed into chat — and this agent turns that into a written SOP: numbered steps, in order, searchable. The draft stops in a review queue with a named person on it. Nothing gets published until that person approves it, and nothing already in your wiki gets overwritten to make room for it.
Yours outright — no required subscription.
Every agent is built for your business — your systems, your approval chain, your way of working. We scope it on the call.
If this sounds too technical, don't worry. We take care of everything for you.
- Private VPS deployment. This agent runs on a server that belongs to you, not a shared cloud tenant. Your process documentation, your recordings, your transcripts — none of it sits next to anybody else's.
- A technician agent alongside it. Every hire ships as a pair: the SOP agent that drafts and publishes documentation, and a VPS-technician agent on the same box that keeps it healthy — patched, backed up, monitored. You're not hiring one thing, you're hiring a small team of two.
- Capture from wherever you already explain things. Wired into the recorded walkthrough, the transcript, or the chat thread where the explanation actually happens. You don't have to change how you teach a process in order to get it written down.
- A review queue with a named approver. Every draft lands in front of a person before it goes anywhere. Their name and the approval date go on the published document, so the SOP has an owner instead of an anonymous origin.
- Publishing into the system you already use. SOPs go where your team already looks for documentation, in the structure and tagging you already use. New versions land alongside old ones; nothing existing is deleted, archived, or overwritten.
- A live walkthrough, then 14 days of priority support. We sit down on a call and show you exactly what each agent does and where it's supposed to stop, then stay close for two weeks while your real processes go through it. After that we remove our own access.
- You own everything — the server, the agent, the code it runs on, and every document it produces.
Not sure it fits? Check fit in 90 seconds in the free assessment chat.
Who this is for
- You're running an agency where every client kickoff means someone re-explains the same setup sequence from memory, and the one person who knows the whole thing start to finish is out this week.
- You're a contractor and the job closeout sequence lives in the head of one project manager who has been there eleven years and recently started using the word retirement.
- You have a dozen recorded walkthroughs sitting in a folder called Training that nobody watches, because you can't skim a 22-minute video to find the one step you actually forgot.
- You run a clinic or a front desk that turns over every eight months, new hires learn by shadowing whoever has been there longest, and the steps drift a little further off with each handoff.
How it earns trust
The failure mode of AI in documentation isn't that it gets a step wrong. It's that a wrong SOP looks exactly like a right one — clean, numbered, confident — and a new hire follows it anyway. This agent is built around the opposite bet: every step points back at where it came from.
Every step cites its source
A timestamp in the walkthrough, a line in the transcript, or a link to the thread it was drafted from. So when a step is wrong, you don't argue with the document — you jump to 8:14 in the recording and see whether the agent misread it or the person narrating misspoke.
A name and a date on every published SOP
Drafted, reviewed by a specific person, published on a specific day. And because nothing is ever overwritten, the prior version is still sitting there to diff against, so what changed and when is a question your existing wiki history can answer without the agent's help.
Answers point at documents, not at a black box
When it answers a "how do we…" question, it names the SOP and the section it pulled from. A wrong answer traces to a specific document you can go open and fix, which also tells you the answer will be right for the next person who asks.
None of that makes the agent right — it makes being wrong something you can find, trace, and fix in minutes instead of debating in a thread.
Pairs well with
These share a workflow with this role. Tick any to add them to your setup.
We have SOPs. Nobody reads them and half of them are wrong.
The reading part it can help with: SOPs come out as numbered, searchable steps instead of a video you have to scrub through. The wrong part is honest work. It flags documents that haven't been touched in a while and prompts a review, but that's age, not truth — a process that changed last Tuesday in a doc edited last Tuesday gets no flag. It can tell you what's old. It cannot tell you what's wrong.
I've got a dozen recorded walkthroughs and nothing written down.
That's the input it's built for. It works from the recording or the transcript and drafts the steps in order, with screenshots where the walkthrough shows a screen. What it can't do is invent the parts that never appeared: the phone call, the step someone did from memory, the rule that only applies to accounts over a certain size. If the walkthrough skipped it, the draft marks it as not observed rather than writing a plausible-sounding step.
It's the same five questions in chat every single week.
It answers those from the published SOP and cites the document and section it pulled from, so anyone can check the source in one click. If nothing published covers the question, it says so instead of guessing — and that gap becomes an obvious candidate for the next SOP.
If she leaves, we're in real trouble — it's all in her head.
This gets a lot of it out of her head, but be realistic about what one walkthrough captures. People narrate the clean version of a task and skip the workaround they've quietly been running for six months, so the SOP can come out wrong in exactly the spot where the workaround lives. It also can't tell a one-off from a rule — if the walkthrough includes "and here we called the client, they were upset," that can land as step 5. The review queue exists for that reason: the person who knows the process reads the draft before anyone else does.
What is it never going to do to our documentation?
Four hard stops. It never publishes an SOP a named human hasn't approved. It never deletes, overwrites, or archives an existing document — new versions go alongside the old ones, so your history survives even when the agent gets something wrong. It never fills a gap it didn't observe. And it never certifies a process as correct or compliant: clinical protocols, safety procedures, and licensing requirements get documented, never approved. That sign-off stays with the person whose license and liability it is.
What's covered under the $1,997 setup, and what isn't?
One company, one system of record for documentation, one approval chain. Setup is scoped on a free 15-minute call, and it's typically live about seven days from kickoff. If you keep documentation in two different systems, or different departments need genuinely different approvers, that's a scoped conversation on the call rather than a price adjustment.
Hiring more than one? A department on tap — the subscription puts a build team behind every request, agent after agent.