What they built
An AI hiring analyst by connecting Claude (via StackOne) to Ashby, enabling instant, context-rich answers to hiring pipeline questions - without manually exporting data or configuring reports.
Why they built it
- Spotted a candidate withdrawal in Slack (candidate dropped out of a take-home task citing time constraints) and suspected it wasn't an isolated case
- Couldn't quickly verify the hunch in Ashby before her next meeting - the data existed but required time to configure reports and manually review individual candidate profiles for context
- As a team of one supporting 40 people, needed a faster way to surface data-backed insights to bring to hiring manager conversations
Tools used
- Ashby - ATS holding all candidate pipeline data and context
- StackOne - integration platform that maps Ashby's data into a standard format and exposes it via a built-in MCP gateway (no code required)
- Claude - AI model used to query the connected Ashby data in natural language
- MCP (Model Context Protocol) - open standard that links Claude to external tools and data sources
How you could build this yourself
-
Create an API key in Ashby
- Go to Admin Settings → Integrations in Ashby
- Generate a new API key and scope it to read-only access (candidate profiles, pipeline data, etc.)
- Save the key immediately in your password manager - it only displays once
-
Connect Ashby to StackOne
- In StackOne, select Ashby from the list of 200+ pre-built connectors
- Paste in your Ashby API key
- Review the list of permitted actions and enable only what you need (e.g. read candidate data; leave write/update actions blank)
-
Connect StackOne's MCP gateway to Claude
- Follow StackOne's step-by-step flow to link the Ashby connector to Claude via the MCP gateway
- No developer or custom code required
- Confirm the connection is live by opening a Claude chat and asking: "Do you have access to my Ashby data?"
-
Start querying in Claude
- Create a dedicated Claude project or labelled chat (e.g. "Ashby Reporting") to keep your hiring queries organised
- Ask natural language questions, e.g.: "For the Senior UI Engineer role, how many candidates attempted the take-home task, how many withdrew, and what reasons did they give?"
- Follow up with prompts like "Summarise and categorise the reasons" to make outputs skimmable and shareable with hiring managers
Access control note: Claude will only surface data that your Ashby login already has permission to see - so access is automatically scoped to your role.
Source
Built by Yasmin Chalmers, Head of People at StackOne, and presented as part of an Open Org "Shipped by HR" session.