The short version
You work with AI assistants all day now, whether that’s Claude, ChatGPT, Gemini, Grok, Copilot, you name it. You also work in the CodeSignal platform to check on your candidate reports, assessment sessions, and interviews. The CodeSignal MCP connects the two, so you can pull a report, send an assessment, or check this week’s interviews right from the AI tool you’re already using.
It’s available today! Here’s what it does and how to use it.
First, what’s an MCP?
Model Context Protocol (MCP) is an open standard that lets AI applications connect to outside systems in a consistent way. Think of it as a universal adapter: instead of every AI tool needing a custom integration for every service, an MCP server exposes a service’s capabilities as tools any compatible assistant can use. The CodeSignal MCP is that adapter for CodeSignal.
What your team can get done
We built the MCP around the questions and tasks that already fill your day. Instead of a list of API endpoints, here’s what it looks like by who’s asking:
For recruiters and hiring managers
- “Show me the results and report for candidate X.”
- “Pull the submitted code for session Y.”
- “Create and send an assessment to these candidates.”
- “Set up a new test from a certified assessment.”
- “Extend the deadline on these sessions and resend the reminders.”
- “Spin up an interview from our standard template.”
- “List all interviews scheduled this week.”
For L&D teams
- “Which certified assessments map to the skills we’re trying to build?”
- “How is the team doing on our AI literacy assessments?” and get scores summarized across sessions.
- “Send a skills check to this cohort and track completion.”
For analysts and I-Os
- “How did candidates do on test Z?” and get scores summarized across sessions.
- “Which certified assessments should we use for this role?”
- “Compare pass rates across these two tests.”
Because your assistant can combine CodeSignal data with everything else it can reach, from your notes to a spreadsheet, these stop being separate tools and start being one workflow.
A real example
In our own testing, we gave an assistant a single prompt: “which assessments are available for evaluating AI literacy skills?”
It searched CodeSignal’s certified library, found 65 pre-validated AI literacy assessments spanning 10 job families, and built an interactive view to explore them by skill and role, complete with a breakdown chart, all without anyone touching the CodeSignal dashboard. That’s the shift: from looking things up to asking for what you need and getting it back ready to use.
What the MCP touches
Under the hood, the CodeSignal MCP exposes CodeSignal’s core hiring and assessment workflows as tools. Both reading and doing:
| Area | Read | Write |
| Test sessions | List sessions Get a session’s details and report (including submitted solutions and answers) Get standardized test sessions Look up ATS test sessions | Create and send, resend, reactivate, delete, archive/unarchive, edit duration or expiration, set reminders, grade or mark as graded |
| Live interviews | List interviews Get interview details | Create, edit, or delete an interview |
| Interview templates | List templates Get template details | Create, edit, or delete a template |
| People and organization | Get a user Get your company info List company roles Get a specific role | N/A |
Setup takes about a minute
Implementation is like any other MCP. It authenticates with an API key that you pass as a header, so it works with any AI tool whose MCP support allows custom headers on a remote server. No middleware or bespoke integration needed to maintain!
- Create an API key. In your CodeSignal dashboard, go to Integrations, then API Keys, and create one. You’ll need admin privileges. Don’t see API key settings? Contact support@codesignal.com to enable API access for your organization.
- Add the CodeSignal MCP server to your AI tool. Point your client at the CodeSignal MCP endpoint: https://codesignal.com/api/mcp
- Paste in your API key to authenticate, and start prompting.
Full instructions are on Github.
Your data stays scoped to your key
Quick but important note: The MCP can only see and do what your API key is permitted to, nothing more. In the AI literacy example above, the assistant could search CodeSignal’s certified assessment library but not the company’s own custom tests, because that key didn’t have read permission for them. Access follows your permissions by design.
Try it
If your team runs on CodeSignal, the MCP puts that work right where you already are.