Point Claude at your ATS and let it work
Jwalix exposes a native Model Context Protocol surface, so an MCP-aware client — Claude, or an agent you built — can search candidates, open jobs, create submissions and schedule interviews directly, with no bespoke integration to write and no permission model to reimplement.
Your recruiting data, where you are already working
MCP is the emerging standard for letting an AI client use an external system. Supporting it natively means you are not waiting on an integration to be built for whichever assistant your team adopts.
Bring your own assistant
Use the AI client your team already prefers, against live recruiting data, instead of switching tools to get AI help.
The same thirteen tools
External clients get exactly the actions the in-app assistant has. Both read one definition, so neither surface can quietly fall behind the other.
Automate past the interface
Script the recurring work — a Monday pipeline brief, bulk submission hygiene — as tool calls rather than clicks.
A key that authenticates, and nothing more
The obvious worry about handing an AI client a key is that the key becomes a way around your permissions. Here it is not one.
It runs as the person who made it
A key authenticates as its creator and inherits that account's permissions. It cannot reach a module or a record its owner could not open in the interface.
Shown once, revocable always
Generate a key from your profile and copy it there and then — only a hash is stored. Revoke it whenever you like and it stops working immediately.
Visible use
Keys are listed with a label and a last-used timestamp, so a forgotten key is something you can spot and retire rather than wonder about.
Tenant-scoped like everything else
A key resolves to one workspace. The multi-tenant isolation that separates your data from every other customer applies identically here.
How it is built
The MCP server runs as a separate process that holds no database credentials and no business logic — it discovers the available tools over HTTP and forwards calls. Every permission decision is still made by the platform itself, which is what keeps this an access route rather than a second door. In short: a key authenticates, it does not widen access.
Questions about MCP
The Model Context Protocol is an open standard for connecting an AI client to an external system's data and actions. A client that speaks it can use any MCP server without a purpose-built integration.
Any MCP-aware client, including Claude Desktop and agents you write yourself. Because the protocol is the contract, support does not depend on us shipping a per-client integration.
No. It authenticates as the user who generated it, and every tool call is then checked against that user's own module permissions — the same checks the in-app assistant goes through.
Revoke it from your profile and it stops working at once. Only a hash of the secret is stored, and the last-used timestamp helps you see whether it was used before you caught it.
Yes — through the permissions of the account the key belongs to. Generating the key from a narrowly-scoped user is the straightforward way to keep an automation tightly bounded.
Connect your assistant to a real workspace
We'll generate a key on the call and run live tool calls against your own data.