Skip to content

For MSPs and internal IT

Do the job once. Get the SOP your team follows and your AI can read.

The escalation on your board today was fixed once already, and the fix, with the reason behind it, lives in one technician’s head.

Record the fix in the Recording App, in the PSA, the vendor portal or the client’s own tools. Steptree writes the runbook with the steps, the screenshots and the why.

Approve it, then serve it over MCP. The next tech asks Claude Code, Cursor or GitHub Copilot in VS Code and gets your runbook back.

Free plan: 300 AI credits a month and two recordings, no card. Claude can connect on any plan.

The mechanism

Nothing is published before you approve the runbook

  1. You record the fix in the Recording App

    The Recording App records any app on Windows or Mac: the PSA, the RMM, the vendor console with no API and the terminal. Flag a judgement call as you make it, and say why.

    Password fields are never recorded, and emails and card numbers are stripped from typed text.

  2. Steptree writes the runbook, and you approve it

    The goal, the steps with screenshots under them, the values that change per client and the calls you flagged. Edit it like a doc. It stays a draft until you publish it.

  3. The next tech reads it, or asks their AI

    Share it with the team that works that client, or connect Claude Code, Cursor, Codex or GitHub Copilot over MCP. They search your runbooks and answer from them.

    MCP is on every plan, including free.

The artefact

Read the runbook before you believe anything else here

Every runbook opens with the goal your technician stated: which console, which tab, and the field that stays greyed out until you set another one.

A runbook your next hire could follow cold

The values that change per client sit in braces: the tenant, the licence pool, the mailbox. The decision points your technician flagged sit with the reason for each, and the screenshots sit under the steps.

Where your techs already work

Your runbooks, inside the editor your techs have open

Connected once over MCP, not pasted into a chat every morning.
Claude Code, Cursor, Codex or GitHub Copilot
Each has a setup guide. A tech asks in the editor they already have open and gets your runbook back, not a guess.
Each tech reads only what they can open
The MCP key is personal and expires. Results follow that person's own access, so a client's runbooks stay with the team that works it.
Read-only unless you allow edits
Mint a read-only key for the queue, or allow edits for whoever keeps the runbooks current. Nothing publishes itself.
A runbook can also be a skill file
Turn any runbook into a SKILL.md your AI tools can load. Plain text, yours to keep.

One escalation proves the runbook holds up. What a subscription buys is room to record every fix that repeats, so the next tech closes the ticket from your runbook instead of waiting on whoever fixed it before.

Questions

How this differs from what you already pay for

How is this different from the documentation tool we already have?
Nobody types the runbook. Steptree writes it from the fix as your technician actually worked it, with the screenshots and the calls they flagged. So it matches the screen, not somebody's memory of it.
Can our techs read runbooks from Claude Code or VS Code?
Yes. Steptree serves your library over MCP on every plan. Claude Code, Codex, Cursor and GitHub Copilot in VS Code each have a setup guide.
How do we get the runbooks out, and what if we leave?
There is no runbook export to PDF or Word today. What there is: any runbook can become a SKILL.md file you download, and MCP on every plan reads your runbooks and recordings straight out of Steptree. Each person can also download their own data from Settings.
How do you keep one client's runbooks away from another's?
Runbooks belong to teams, so a client's runbook shared with its team is visible to that team and your admins. Your AI reads through the same access, because an MCP key follows its owner.
Does this work for internal IT, not just an MSP?
Yes, with the client column removed. The fixes only one person knows, the console with no API, and the new starter still learning where things are. Teams map to your own departments instead.
What does it cost to find out if this works?
Nothing. The free plan has 300 AI credits a month and two recordings, no card, and MCP is included. That is enough to record one escalation and read back the runbook. Paying is what documents every fix that repeats, not just the one you tested.

Next

Where to go from here

Work one escalation, and read what comes back

Record the fix that keeps coming back, and read the runbook Steptree writes from it. Free gives you 300 AI credits a month and two recordings, no card: enough to prove it before the ticket repeats.

Someone at a wall of sticky notes talking a seated team through a process.