What teams use Toolspoke for
Each of these is a page rather than a paragraph, because the answer to each one is a mechanism you can go and read.
Access that follows your team
Who is allowed to call what?
A project decides which of a toolkit's actions it allows, and an agent key gets a slice of that. A key never reaches past what its owner can reach today, and both checks run when the call arrives rather than when the menu is drawn.
Read the pageAccurate tool choice in a large catalog
Why does my agent keep picking the wrong one?
Tool-selection accuracy falls away once a client holds more than a few dozen tools. Toolspoke sends a searchable index instead of every schema, so the model chooses from a shortlist it can reason about.
Read the pageA gateway on your own infrastructure
Can this run inside our network?
Docker Compose and PostgreSQL 17. Two secrets to set, everything else optional, and no dependency on our servers once it is running. Billing is not configured unless you configure it.
Read the pageA record of every agent call
What exactly happened, and who is it attached to?
Every call writes a row: the person behind the key, the client that made it, the arguments, the result, and where the time went. Including the limits, which are stated rather than glossed over.
Read the page
If what you are trying to do is not one of these, the integration catalog and the documentation are the fastest way to find out whether it fits.