Codex + InterviewFlowAI

Work with Candidates and AI Interviewers in Codex

One codex mcp add command for both the CLI and Codex Desktop
Filter candidates by score, export CSVs, and draft AI Interviewers
Every change and candidate email waits for your confirmation

InterviewFlowAI is the AI interview platform Codex connects to through MCP. Please keep a human reviewing every change. Part of the InterviewFlowAI AI agents guide.

Recruiter asking an AI assistant connected to InterviewFlowAI a hiring question

Illustrative example

How to install InterviewFlowAI in Codex

Three steps, including a browser sign-in. You need an InterviewFlowAI account on a paid plan or the 7-day free trial, and no API key.

1Installa few minutes

Add InterviewFlowAI to Codex

Run this command in your terminal:

Terminal

# Add the server (the CLI and Codex Desktop share this config)
codex mcp add interviewflowai --url https://api.interviewflowai.com/mcp

# Sign in
codex mcp login interviewflowai

# Verify, then restart Codex
codex mcp get interviewflowai

Or add the server by hand. Open or create the Codex configuration file (~/.codex/config.toml on macOS and Linux, $HOME\.codex\config.toml in Windows PowerShell, or %USERPROFILE%\.codex\config.toml in Command Prompt). The Codex CLI and Codex Desktop share this file.

config.toml

[mcp_servers.interviewflowai]
url = "https://api.interviewflowai.com/mcp"

Sign in with your InterviewFlowAI login (email, Google, or LinkedIn) and click Allow access. Or paste the setup prompt from Settings → MCP and AI and let the agent add the server for you. Details in the MCP documentation.

2Asklive in InterviewFlowAI

Ask Codex about your pipeline

Type the request in plain English. Codex asks InterviewFlowAI for the data and answers from your workspace.

Show me the top 3 candidates by interview score for the Support Lead Interviewer.

8 candidates completed the interview. Here are the top 3:

Priya Raghunathan92
Daniel Okafor89
Mei Lin Chow86

via InterviewFlowAI · example data

Also try

Which candidates scored above 70?
Export the Support Lead candidates
Clone the Sales Interviewer

Add the InterviewFlowAI serverRun this command in your terminal: codex mcp add interviewflowai --url https://api.interviewflowai.com/mcp
Sign inRun this command, then approve the connection in the browser window that opens. In Codex Desktop, you can instead go to Settings → MCP Servers and select Authenticate for interviewflowai. codex mcp login interviewflowai
Verify and restartRun codex mcp get interviewflowai and confirm the server is enabled. Restart Codex so it loads the InterviewFlowAI tools.
Ask your first questionStart read-only: “Show me the candidates who completed the interview.” Any change starts as a preview you confirm.

One connection gives Codex your whole hiring pipeline

Connect once and Codex can reach every AI Interviewer, candidate, and result in your workspace, with the same access as your InterviewFlowAI role.

Codex

AI InterviewersCandidatesInterview scoresResume feedbackRole playsHiring stagesRemindersInvitesNotesCSV export

AI Interviewers
Candidates
Interview scores
Resume feedback
Role plays
Hiring stages
Reminders
Invites
Notes
CSV export

Which way should I connect InterviewFlowAI to Codex?

Use the first option unless you have a reason not to. Every route reaches the same InterviewFlowAI MCP server at https://api.interviewflowai.com/mcp.

Method	How you connect	Sign-in	Best for
codex mcp add (recommended)	`codex mcp add interviewflowai --url <server URL>`	Browser sign-in with codex mcp login	Most Codex users
config.toml by hand	`[mcp_servers.interviewflowai] with the server url`	Browser sign-in with codex mcp login	Managing config files directly
Setup prompt	`Settings → MCP and AI → Copy setup prompt, then paste it into Codex`	Browser sign-in, started for you	Letting Codex do the setup
API key	`Send Authorization: Bearer ifk_… with each request`	API key that acts as the Owner who created it	Scripts and tools that can’t open a browser

The same server works with other agents too. See InterviewFlowAI MCP and the AI agents overview.

Why use Codex for recruiting?

Codex already sits where you work. With InterviewFlowAI connected, a question about your pipeline gets an answer from your real candidates and interviews instead of a generic one.

Ranked shortlists

Sort and filter candidates by interview, resume, or assessment score, and get a ranked list back in one request.

Interview and resume detail

Read interview results, resume feedback, and each candidate’s timeline of emails, reminders, and stage changes.

Role play results

List your role plays and read each candidate’s graded result alongside their AI interview.

Pipeline actions

Owners can star candidates, move them between hiring stages, and update notes and custom fields.

Reminders and invites

Owners can remind unfinished candidates, invite new ones, give extra time, and send role plays.

AI Interviewers from a job description

Owners can draft an AI Interviewer from a job description, or create and clone one, then fine-tune it in InterviewFlowAI.

What is Codex?

Codex is OpenAI’s coding agent, available as the Codex CLI and the Codex Desktop app. Both read MCP servers from the same config.toml, so adding InterviewFlowAI once makes it available in each.

Codex or ChatGPT?

Use Codex if you work in the terminal or Codex Desktop and want files and exports in your project. Use ChatGPT for chat-style candidate review. Both use the same InterviewFlowAI server URL.

Client	Where it runs	Who it’s for	How you add InterviewFlowAI
Codex	Terminal and Codex Desktop	Teams already using Codex	codex mcp add command or ~/.codex/config.toml (CLI and Codex Desktop)
ChatGPT	ChatGPT	Recruiting teams	Connector in Settings → Apps & Connectors

One config for CLI and app

The Codex CLI and Codex Desktop share config.toml, so you add InterviewFlowAI once.

Job descriptions from your files

Point Codex at a job description in your project and turn it into an AI Interviewer draft.

API keys for scripts

For a script or tool that can’t open a browser to sign in, use an InterviewFlowAI API key instead of OAuth.

How InterviewFlowAI works with Codex

Add InterviewFlowAI as an MCP server and Codex sees the InterviewFlowAI tools the moment you sign in. Your role decides which ones: Viewers get the read tools, and Owners also get the tools that make changes.

codex mcp add command or ~/.codex/config.toml (CLI and Codex Desktop)
Sign in with your InterviewFlowAI login, with no API key to manage
Viewers stay read-only; Owners can act after confirming a preview
Up to 60 requests per minute across every assistant you connect
Disconnect any time from Settings → Connected apps

Install in Codex

# Add the server (the CLI and Codex Desktop share this config)
codex mcp add interviewflowai --url https://api.interviewflowai.com/mcp

# Sign in
codex mcp login interviewflowai

# Verify, then restart Codex
codex mcp get interviewflowai

How do I review candidates with Codex?

Ask in plain English. Codex finds the right Interviewer, reads the results, and answers. When you ask for a change, it shows you a preview first and waits for your yes.

Just ask

Read-only, safe to start with

Ask about any Interviewer, candidate, or score. Nothing changes in your workspace.

Codex · example

> Which candidates finished the Support Lead interview
  this week? Rank them by interview score.

Codex: 8 candidates completed it. Top 3:
  1. Priya Raghunathan    92
  2. Daniel Okafor        89
  3. Mei Lin Chow         86
  Want their interview summaries?

Confirm before it acts

Owners only, every change previewed

Starring, stage moves, reminders, and invites start as a preview listing who is affected and who is skipped.

Codex · example

> Star the top 3 and move them to the Onsite stage.

Codex: Here is the preview.
  Star            Priya, Daniel, Mei Lin
  Move to Onsite  3 candidates
  Skipped         none
  Confirm? The preview expires in 10 minutes.

> Yes, confirm.

Codex: Done. 3 candidates starred and moved to Onsite.

What can you ask Codex to do in InterviewFlowAI?

Anything your role allows: Interviewers, candidates, results, stages, reminders, and exports. Here are typical requests and what happens with each one.

Filter by score01

List candidates in the Account Executive Interviewer with an interview score of 70 or above and an assessment score above 60.

Sort and filter by interview, resume, or assessment score in one request.

Export for reporting02

Export all candidates for the Support Lead Interviewer as a CSV.

The download link works for 15 minutes; within the hour you can ask for a fresh one.

Give more time03

Give the shortlisted candidates 5 more days to finish their interview.

Only shortlisted candidates can get more time, from 1 to 60 days. You confirm the preview first.

Draft an Interviewer04

Draft an Interviewer from the job description in ./roles/sdr.md.

Codex reads the file and drafts the role and questions for you to review before it is created.

What can Codex access?

Codex gets the same access as your InterviewFlowAI role, so it can never do more than your own account. Viewers’ assistants never see the tools that make changes.

Owners and Viewers

Read and search

List and search AI Interviewers, and sort and filter candidates by interview, resume, or assessment score
Read candidate details, interview results, resume feedback, and timelines
List your role plays and read each candidate’s graded role play results
List your custom hiring stages and export candidates to a CSV file

Owners only, after confirming a preview

Act on candidates and Interviewers

Star candidates, move them to a hiring stage, and update notes and custom fields
Send reminders, invite new candidates to an interview, and give candidates more time
Invite candidates to a role play
Draft an AI Interviewer from a job description, create or clone one, and change its visibility or archive it

60Requests per minute, shared across every assistant you connect

50Candidates per action. Reminders can go to up to 100 at a time

10 minBefore a change preview expires and the assistant shows a fresh one

15 minHow long a CSV export link works. Anyone with the link can download it

Resumes, interview answers, and other candidate content can contain instructions aimed at AI assistants. If Codex proposes a change or an email you didn’t ask for, say no, and keep its own approval prompt on for InterviewFlowAI tools. Scripts that can’t open a browser can use an InterviewFlowAI API key instead.

Frequently asked questions

Can’t find an answer? Read the MCP documentation or email support@interviewflowai.com.

How do I connect InterviewFlowAI to Codex?

Add the InterviewFlowAI server, sign in, then verify and restart. The server URL is https://api.interviewflowai.com/mcp. Setup takes a few minutes, and you sign in with your usual InterviewFlowAI login.

Do I need an API key to use InterviewFlowAI with Codex?

No. Codex signs in with OAuth: you approve the connection with your InterviewFlowAI email and password, Google, or LinkedIn login. You only need an API key for a script or tool that can’t open a browser to sign in. An API key acts as the Owner who created it, including write actions, so keep it private.

What can Codex do with InterviewFlowAI?

Everyone can list and search AI Interviewers, filter candidates by interview, resume, or assessment score, read interview results, resume feedback, timelines, and role play results, and export candidates to CSV. Owners can also star candidates, move them between stages, update notes, send reminders and invites, give candidates more time, invite them to role plays, and create or clone AI Interviewers.

Can Codex change candidates or email them without asking me?

No. Before any change, and before any email to a candidate, Codex shows you a preview of what will happen, including any candidates it will skip and why. Nothing happens until you confirm, and a preview expires after 10 minutes. Viewers have read-only access, and their assistant never sees the tools that make changes.

Codex doesn’t see the InterviewFlowAI tools. What should I check?

Run codex mcp get interviewflowai and check the server is enabled. If it is, run codex mcp login interviewflowai again, then restart Codex. Codex only loads new MCP tools after a restart. Also check that your InterviewFlowAI subscription or 7-day trial is active, because an expired account refuses the connection.

Is MCP included in my InterviewFlowAI plan?

Yes. MCP is available on every paid plan and during the 7-day free trial. You can make up to 60 requests per minute, counted across all the assistants you connect.

How do I disconnect Codex?

Open Settings → Connected apps in InterviewFlowAI and click Disconnect next to it. The assistant loses access right away and must ask you again before it can reconnect.

Use InterviewFlowAI with another AI agent

Every agent connects to the same MCP server with your own InterviewFlowAI login.

Start screening with your AI agent today

Connect InterviewFlowAI to Codex once and let it find, review, and act on candidates while you stay in control of every decision.

Start free trial Connect MCP in your workspace

Recruiter reviewing InterviewFlowAI candidates ranked by interview score