Walter answers the phone, reads a live Google Calendar and books the meeting. It runs on any customer's account, with no per-customer code behind it.
Professional AI Agents LLC builds bespoke agents that work alongside a client's existing staff. Every one is built for the business it serves, and every one has to connect to the systems that business already runs.
"With the bespoke solutions, all of it is custom integrated to a business. So if I need to scale, that's been a very hard problem for me to solve. That's been another integration with that client, and then running different API integrations and seeing exactly what their workflow is."
The logic is the part worth building. The connections, the auth and the monitoring around it are the part that takes the time, and they arrived fresh with every new agent and every new customer.
Map the workflow, read the API docs, handle pagination, retries and rate limits. Per client, every time.
An OAuth flow, a callback, somewhere to put the refresh token, and revocation to handle when it expires.
Pick a monitoring architecture, decide how granular to instrument inside the code, then rebuild it per agent.
The demo at the top of this page runs against a single Google Calendar. The hard engineering was finished. Selling it was a different problem, and it is where most internally-built agents stop.
Every buyer's account needs its own consent flow, its own callback, and its own stored grant.
Refresh before expiry, handle revocation, re-consent on scope changes. For every account the agent ever touched.
One path for the builder's own account, one for tenants, and both branches kept in step forever.
The agent calls a tool against a connected account ID. Scalekit resolves the user, pulls a scoped credential, executes the call and returns the result. No token in the agent process, none in the LLM context. Walter uses Google Calendar; the pattern is identical whichever tool the next agent needs.
The agent asks for availability or writes an event, referring to a connected account rather than a credential.
It looks up that customer's connected account and pulls a scoped credential, kept refreshed outside the agent.
Scalekit runs the scoped call against the Google account the customer connected, and returns the result.
The meeting lands on that customer's calendar, with no per-customer code anywhere in the path.
"Because of the Scalekit integrator, I don't actually need to build additional Python code so that somebody else could use it. They could connect their Google Calendar to Walter, and then all of a sudden it actually is a sellable product. Whereas previously, it would've just been my own digital employee. Scalekit has allowed me to sell this."
The same connection layer, whichever account your customer connects, so one agent serves all of them.
We use cookies, so things load fast, we learn what to fix, and you can always reach us on chat
