Build Your Own AI Recruiting Assistant With Claude Code, Even If You're Not a Developer

A recruiting assistant built on Claude Code needs a paid Claude plan, a folder of your own files and a page of written instructions. As of 16 October 2026, Anthropic's pricing page lists Pro at $20 a month billed monthly, and the Free plan does not include Claude Code. You do not write code, but you do own the data handling and the hiring-law risk.

This guide is for recruiters and HR generalists. Anthropic publishes no recruiting benchmarks for the tool, so nothing here promises a time saving. It shows what the product is documented to do, how to set it up, and where the legal limits sit.

What is Claude Code, and can a non-developer use it?

Claude Code is, in Anthropic's words, an agentic coding tool that reads your codebase, edits files, runs commands and integrates with your development tools. It is available in the terminal, in IDEs, in a desktop app and in the browser. A non-developer can use it because you give instructions in plain language.

The catch is that the documentation is written for software teams. Its examples cover tests, bug fixes and pull requests. A recruiting assistant is an application of the same abilities (reading files in a folder, writing new files, following saved instructions) to your own material. Treat that as your design choice, not a vendor-supported recruiting product.

How much does it cost to run a Claude Code recruiting assistant?

Pro is the cheapest plan that includes Claude Code. All figures below are from Anthropic's pricing page as of 16 October 2026, exclude tax, and are subject to change at Anthropic's discretion.

Usage limits apply on every plan, so a heavy resume-processing day may hit a cap. Anthropic's Team plan listing says no model training on your content by default, which matters for candidate data (see below).

What can a Claude Code recruiting assistant actually do?

It can read files you put in a folder, write new files, and follow instructions you save. Anthropic's docs describe MCP as an open standard for connecting Claude Code to outside data sources, with Google Drive, Jira and Slack as examples. Everything in the table below is a suggested use built on those abilities, not a tested result.

For sourcing strings, see our Boolean search guide for recruiters. For a wider view of tools around the assistant, see The Headhunter's Toolkit.

How do you set it up, step by step?

The shortest path uses the desktop app. Anthropic says that after you install it, you sign in and click the Code tab, and the app includes Claude Code so you do not need the CLI separately. A paid subscription is required.

1. Pick a plan. Choose a plan from the table above. For anything involving real candidate data, read the data section below first.
2. Install the desktop app. Download it from the link in Anthropic's docs for macOS or Windows, sign in and open the Code tab. If you prefer the terminal, the docs say a terminal guide shows first-timers how to open one and paste the install command.
3. Create a folder per search. Use plain names, for example one folder for each open role. Keep resumes and notes inside.
4. Open that folder in Claude Code. The tool works on the files in the project directory you point it at.
5. Add a CLAUDE.md file. This is where your rules live (next section).
6. Start in Manual mode. Review every file edit until you trust the output.
7. Correct and save. When you fix the same mistake twice, add the fix to CLAUDE.md.

What should go in the CLAUDE.md file?

CLAUDE.md is a markdown file you add to your project root that Claude Code reads at the start of every session. Anthropic's memory docs say to add to it when Claude makes the same mistake a second time. Write it in plain sentences.

One caution from the same docs: Claude treats these instructions as context, not enforced configuration. A rule like "never send email" is a request, not a lock. Use permission settings for hard limits.

A starter you can copy and edit:

Recruiting assistant rules:
- I am a recruiter. You help me prepare, I make every hiring decision.
- Summarise candidates only against the criteria in scorecard.md.
- Quote the resume line that supports each point. If there is no evidence, write "not shown".
- Never infer age, gender, race, religion, disability, family status or nationality.
- Never rank, score or reject candidates. List evidence only.
- Write outreach as drafts. Do not send anything.
- Keep a short list of anything you were unsure about at the end of each file.

Anthropic also describes skills: markdown files with workflows you can invoke by name, such as a saved "weekly pipeline update" routine. Its guidance is that you do not need to configure everything up front, and most teams add features as specific triggers come up.

Is it safe to put candidate data into Claude Code?

It depends on your plan, so decide before you upload anything. Per Anthropic's data usage page, consumer plans (Free, Pro, Max) can have data used to improve models when that setting is on, including Claude Code use. Commercial plans (Team, Enterprise, API) are not used for training unless the customer opts in to a program such as the Development Partner Program.

These are Anthropic's own policy statements, as of 16 October 2026. Two more points from the same page. Claude Code stores session transcripts locally in plaintext for 30 days by default, so protect your laptop. And transcripts you send with /feedback are retained for 5 years, so never attach one that contains candidate details.

On control: Anthropic's permission modes set what Claude can do without asking. In Manual mode, Claude stops before most actions that edit files, run shell commands or reach the network, and the docs recommend it for sensitive work. The docs also say that from Claude Code v2.1.283, auto mode is the built-in starting mode for interactive terminal and VS Code sessions, so check the mode indicator and switch to Manual yourself.

Which AI hiring laws apply to a recruiting assistant?

Mostly the ones that cover tools that influence who gets hired. Drafting emails sits outside most of them. Ranking or rejecting candidates does not. The table summarises three, and none of it is legal advice.

On the New York test, according to DLA Piper, a tool "substantially assists" a decision when the employer relies on its output exclusively, weighs it above any other criterion, or uses it to overrule human judgment. That is why the starter rules above tell the assistant to list evidence and never rank. It reduces exposure, but it does not prove you are outside the law. Staffing agencies are named in the NYC rule, so agency owners should take it seriously.

If you want a packaged product with vendor-supplied audits instead, compare options in Paradox vs HireVue and the ATS features staffing agencies use. For a free place to keep the official candidate record, see the best free ATS for your first 10 hires.

What's still unclear?

Several points are open, and the weakest part of this approach is the lack of evidence.

• No published recruiting results. Anthropic's docs show developer workflows. There is no independent measurement of recruiter time saved or screening quality with Claude Code that this guide can cite, so any savings you expect are your own estimate to test.
• Accuracy of summaries. A summary can miss or misstate a resume line. The starter rules ask for quoted evidence so you can check, but you still have to check.
• Colorado wording. McDermott describes SB 26-189 as scaling back the earlier law, while AI Policy Tracker describes a repeal and replacement, and notes its record comes from secondary reporting pending review of the enrolled bill. The 1 January 2027 date rests on that one tracker.
• Whether your setup is an AEDT. The NYC test turns on how much weight you give the output. That is a fact question for your counsel.
• Plans change. Anthropic says prices and plans are subject to change, and Claude Code defaults have already shifted by version.
• ATS connections. Anthropic's examples of MCP connections do not include recruiting systems, so integration with your ATS is unconfirmed.