--- title: "AI Sourcing With Claude and Leonar (MCP)" description: "Use Claude to search, rank and qualify candidates in Leonar's own database through MCP. LinkedIn Recruiter stays a separate step you run yourself." url: "https://www.leonar.app/blog/ai-sourcing-with-claude-and-leonar/" lang: "en" alternates: fr: "https://www.leonar.app/fr/blog/ai-sourcing-with-claude-and-leonar/" de: "https://www.leonar.app/de/blog/ai-sourcing-with-claude-and-leonar/" image: "https://www.leonar.app/og/blog/ai-sourcing-with-claude-and-leonar-en.png" tokens: 4786 --- 1. [Home](https://www.leonar.app/) 2. [Blog](https://www.leonar.app/blog/) 3. [Sourcing](https://www.leonar.app/blog/category/sourcing/) 4. AI Sourcing With Claude and Leonar: How It Works Sourcing July 3, 2026 11 min read # AI Sourcing With Claude and Leonar: How It Works Use Claude to search, rank and qualify candidates in Leonar's own database through MCP. LinkedIn Recruiter stays a separate step you run yourself. [![Pierre-Alexis Ardon](https://www.leonar.app/_astro/pierre-alexis-ardon.ByBiMn-t_1eJ10c.webp)](https://www.leonar.app/authors/pierre-alexis-ardon/) [Pierre-Alexis Ardon](https://www.leonar.app/authors/pierre-alexis-ardon/) Co-founder Updated September 2, 2026 ![Adrien Tedjirian](https://www.leonar.app/_astro/adrien-tedjirian.BbnmZr4f_18wWaF.webp) ![Dolihane Feddag](https://www.leonar.app/_astro/dolihane-feddag.DkmcuiI8_uRaLP.webp) ![Louis de Froment](https://www.leonar.app/_astro/louis-de-froment.Y5ImtxTq_28zrXp.webp) Trusted by 400+ recruiting agencies 4.9 Rated on G2 ![Cover card: AI sourcing with Claude and Leonar through MCP, no browser bots](https://www.leonar.app/_astro/ai-sourcing-with-claude-and-leonar.DlfS3aJK_Z1FWPon.webp) Ready to take action? Discover Leonar in a few minutes. [See pricing](https://www.leonar.app/pricing/) - 7-day free trial - Full feature access - Cancel anytime Summarize with AI Choose an AI [ChatGPT](https://chat.openai.com/?q=Summarize%20this%20article%20about%20recruiting%20from%20Leonar%3A%20) [Perplexity](https://www.perplexity.ai/search/new?q=Summarize%20this%20article%20about%20recruiting%20from%20Leonar%3A%20) [Claude](https://claude.ai/new?q=Summarize%20this%20article%20about%20recruiting%20from%20Leonar%3A%20) [Mistral](https://chat.mistral.ai/chat?q=Summarize%20this%20article%20about%20recruiting%20from%20Leonar%3A%20) Table of contents - [Quick answer](#quick-answer) - [Why recruiters want Claude in the sourcing loop](#why-recruiters-want-claude-in-the-sourcing-loop) - [Why Claude cannot work inside LinkedIn Recruiter](#why-claude-cannot-work-inside-linkedin-recruiter) - [The obvious method that breaks: giving Claude a browser](#the-obvious-method-that-breaks-giving-claude-a-browser) - [Where it falls apart for real sourcing](#where-it-falls-apart-for-real-sourcing) - [The clean method: give Claude a structured tool instead](#the-clean-method-give-claude-a-structured-tool-instead) - [Step by step: set up AI-assisted sourcing with Claude](#step-by-step-set-up-ai-assisted-sourcing-with-claude) - [Beyond the search: let Claude qualify the shortlist](#beyond-the-search-let-claude-qualify-the-shortlist) - [When a browser agent is actually fine, and when it is not](#when-a-browser-agent-is-actually-fine-and-when-it-is-not) - [FAQ: AI sourcing with Claude and Leonar](#faq-ai-sourcing-with-claude-and-leonar) - [Where LinkedIn Recruiter fits next to Claude and Leonar](#where-linkedin-recruiter-fits-next-to-claude-and-leonar) - [A controlled workflow](#a-controlled-workflow) ## Quick answer Claude can help you source candidates, but not inside LinkedIn Recruiter. The clean path is a structured tool layer: Claude reasons about candidates in Leonar’s own licensed database through MCP, Leonar handles the search, the ranking, the enrichment and the pipeline updates, and anything you do in LinkedIn Recruiter stays a separate step you run yourself. Yes, you can use Claude for sourcing. But probably not the way you first tried. Most people point Claude at a browser and let it click around a recruiting site. It looks clever in a demo. For real sourcing, it falls apart fast, and on LinkedIn it is not allowed. The cleaner path is to give Claude a structured tool that exposes a candidate database and a pipeline, then let Claude do what it is genuinely good at: reasoning about profiles. This guide walks through both methods. You will see why the obvious one breaks, and how to set up the reliable one step by step, using the same prompts a recruiter would actually type. ## Why recruiters want Claude in the sourcing loop Sourcing is the part of recruiting that eats the most time for the least reward when done by hand. You set up filters, scroll through profiles, read the same career histories over and over, and compare experiences one by one. None of that is hard. It is just slow and repetitive. Claude is strong at exactly this kind of reasoning. It can read a profile, summarize a career, explain why someone fits a brief, and rank a list of candidates against your criteria. Ask it to compare two engineers and it will give you a clear, structured answer in seconds. There is one honest limit, and every guide about Claude for recruiters admits it. Claude cannot find or judge candidates from data it does not hold. On its own, it has no talent database. That gap is the whole reason this article exists. If you want to understand the broader category, our explainer on the [AI sourcing agent](https://www.leonar.app/blog/what-is-ai-sourcing-agent/) covers how these tools are meant to work end to end. ## Why Claude cannot work inside LinkedIn Recruiter Here is the real situation in two sentences. The intelligence lives with Claude. The profiles you pay for in [LinkedIn Recruiter](https://www.leonar.app/blog/how-to-use-linkedin-recruiter/) live inside LinkedIn, and LinkedIn does not offer an open way for an outside assistant to plug in. That is not a gap a workaround should fill. LinkedIn’s User Agreement prohibits third-party software and bots that automate activity on LinkedIn or copy its data. So the question is not “how do I get Claude into Recruiter”. It is “where can Claude work on candidate data legitimately”. The answer is a database you are licensed to use, and a workflow where the Recruiter steps stay yours. ## The obvious method that breaks: giving Claude a browser The first idea, and the most common one, is to connect Claude to a browser. Claude looks at the screen, clicks, reads profiles, and tries to act like a human user moving through the interface. The appeal is easy to understand. It is simple to grasp and simple to test. In a few clicks you can launch a demo and watch Claude navigate a real page. For a quick show-and-tell, it works. ### Where it falls apart for real sourcing Push it toward actual volume and the cracks show quickly. Four problems in particular. First, on LinkedIn it is outside the rules. A bot driving your LinkedIn account is exactly what the automated-activity policy prohibits, whatever the assistant behind it. Second, it burns tokens. Claude has to re-read the screen, the pages, and the profiles, sometimes several times over. What feels cheap on two profiles gets expensive fast on two hundred. Third, it is fragile. A pop-up, a slow load, or a small change in the interface can break the whole run. You end up babysitting a workflow that was supposed to save you time. Fourth, it does not scale. For a handful of profiles, fine. To build a real shortlist or run a search at volume, screen-clicking is the wrong model. If you need candidate data out of Recruiter, use [LinkedIn’s own export options](https://www.leonar.app/blog/how-to-export-candidates-from-linkedin-recruiter/) instead. So connecting Claude to a browser is easy. If your goal is to source at any real volume, it is not the clean method. ## The clean method: give Claude a structured tool instead The better approach is to stop asking Claude to drive the screen. You give it access to a structured tool instead. This is where Leonar fits in. It exposes an MCP, so Claude can call its actions directly from its own environment: searching Leonar’s own licensed profile database, ranking candidates against a brief, and writing the results to your CRM. MCP, the Model Context Protocol, is simply a standard way for an assistant to use a tool’s actions instead of clicking pixels on a page. In practice, Claude never touches a LinkedIn page. It receives your request in plain language, uses the available actions through the connection, and Leonar runs the search on its own database. Claude reasons, the tool executes, and LinkedIn is not part of the loop. Here is how the two methods compare on the things that actually matter for sourcing. | What matters | Claude driving a browser | Claude using Leonar's MCP | | --- | --- | --- | | Where the data lives | Whatever site is on screen | Leonar's own licensed database | | Reliability | Breaks on pop-ups, slow loads, UI changes | Stable, runs on defined actions | | Token cost | High, re-reads screens and profiles | Low, no screen reading | | Scale | Fine for a few profiles only | Handles real shortlists and volume | | Maintenance | Needs babysitting | Repeatable, nothing to babysit | | Setup | Very quick to demo | A few connection steps, then repeatable | If you want the deeper technical picture, our guide on [how MCP connects any AI to your recruiting stack](https://www.leonar.app/blog/connect-ai-agents-recruiting-stack-mcp-api/) explains the connection layer in full. ## Step by step: set up AI-assisted sourcing with Claude Here is the full workflow, from a cold start to your first prompt. Four steps. 1. **Create a Leonar account.** There is a free 7-day trial with no credit card required, so you can test the whole flow before committing. It is the layer between Claude and your candidate data, so it needs to be set up first. 2. **Set up your workspace.** In the dashboard, create the project you are sourcing for and import any candidate files you already have. That pool, plus Leonar’s own database, is what Claude works with. 3. **Connect the MCP inside Claude.** This makes the tool’s actions available to Claude. Once it is linked, Claude can read a request in plain language and call the right actions. This is what replaces the whole idea of Claude clicking around a browser. 4. **Run your first sourcing prompt.** With the tools available, describe your need the way you would to a colleague. You do not have to translate your need into filters yourself. You just say what you want. For example: > “Find Java software engineers based in Lyon with at least three years of experience, and build me a first shortlist of relevant profiles.” Claude reads the request, then uses the connection to run the search across Leonar’s database and your own records. No Boolean strings by hand, no clicking through filter menus. If you still like to control the exact query when you search in Recruiter yourself, our guide to [LinkedIn Recruiter Boolean search](https://www.leonar.app/blog/linkedin-recruiter-boolean-search/) is worth a read. ## Beyond the search: let Claude qualify the shortlist Getting a list back is only half the value. The real payoff is that Claude can help you make sense of that list. Once the results come in, you can ask it to compare candidates, prioritize the ones that look strongest, or explain why one profile matches the brief better than another. It reasons over the records, so you get judgment, not just names. Try a follow-up like this: > “Rank these profiles by relevance and tell me quickly why the top three are the best.” That is the moment the difference becomes clear. You are no longer asking software to click. You are using Claude to reason about your sourcing, while Leonar runs the actual search on its own database and the LinkedIn Recruiter steps stay in your own hands. For a broader view of how this fits a full pipeline, see our [step-by-step AI sourcing workflow](https://www.leonar.app/blog/source-candidates-ai-agents-workflow/). ## When a browser agent is actually fine, and when it is not None of this means browser control is useless. It has its place, on sites and tools that allow it. If you want to test an idea on your own internal tools, or run a quick one-off in a sandbox, letting an assistant drive the screen is fine. It is fast to set up and you are not depending on it for anything serious. The moment you want repeatability, volume, or a real shortlist you can hand to a hiring manager, that model stops holding up. And on LinkedIn, it is not an option at all. That is when a structured connection earns its keep. Match the method to the job, and you avoid a fragile setup that looks good in a demo and fails on Monday morning. ## FAQ: AI sourcing with Claude and Leonar **Can Claude access LinkedIn Recruiter directly?** No. Claude has no access to LinkedIn Recruiter data, and Leonar does not give it any. Claude works on candidate records in Leonar’s own database through MCP. Searching and messaging in Recruiter stay with you, in your own account. **Is using an AI bot on LinkedIn Recruiter against the terms?** Yes. LinkedIn’s User Agreement prohibits third-party software and bots that automate activity on LinkedIn, and signing in with your own account does not change that. The clean posture is to let Claude work on a database you are licensed to use, and to perform every Recruiter step yourself. **What is MCP, and why does it matter for sourcing?** MCP, the Model Context Protocol, is a standard way for an assistant like Claude to use a tool’s actions as if they were built-in commands. Instead of reading a screen pixel by pixel, Claude calls a defined action, such as running a search in Leonar, and gets a clean result back. For sourcing, that means less token waste, fewer breakages, and results you can actually build on. **Do I still need to write Boolean strings by hand?** No. You describe your need in plain language, like “senior data engineers in Berlin open to hybrid work”, and Leonar turns that into a structured search on its database. If you also want to run the same search in Recruiter, Claude can draft the Boolean string for you to paste in yourself. **How is this different from a Chrome extension or scraper?** A scraper copies what it sees on a web page. That is fragile, and on LinkedIn it is outside the User Agreement. Leonar’s MCP does not read any web page. It calls defined actions on Leonar’s own database, which is more reliable, cheaper to run, and steady enough for real shortlists. Leonar’s Chrome extension is different again: it creates a candidate record from the details you confirm, one profile at a time, and only when you decide to. **Which AI assistants work with the Leonar MCP?** Any MCP-compatible client can connect, including Claude Desktop and other assistants that support the protocol. This guide focuses on Claude because it is strong at reasoning over profiles, but the connection layer itself is not tied to a single assistant. ## Where LinkedIn Recruiter fits next to Claude and Leonar The setup is simple to summarize. LinkedIn Recruiter stays part of your toolkit, and you run its searches and messages yourself, in your own account. Leonar handles the search on its own licensed database, the enrichment and the pipeline updates. Claude becomes the interface you give instructions to, and the brain that qualifies the results. That is a much cleaner arrangement than asking Claude to pilot a browser, because everything runs through structured actions built for sourcing. It is also how you move from manual searching to a workflow that does most of the heavy lifting for you, while you keep the final call on which profiles matter. If you want to try it on your own pool, start the free trial, connect the MCP, and [add the candidates you decide to keep with the Chrome extension](https://www.leonar.app/features/chrome-extension-sourcing/) alongside your first search. ## A controlled workflow Claude helps analyze and prepare; you stay the one who initiates any action in your own accounts. Centralize multichannel outreach, prepare follow-ups, and keep control of the actions available in each account. ## Frequently asked questions ### Can Claude access LinkedIn Recruiter directly? No. Claude has no access to LinkedIn Recruiter, and Leonar does not give it one. Claude works on candidate records in Leonar's own licensed database through MCP. Anything you do in LinkedIn Recruiter stays a separate step you perform in your own account. ### Why not let Claude use a browser bot? Browser bots are fragile, token-heavy and hard to maintain, and driving LinkedIn with a bot is not allowed under LinkedIn's User Agreement. A structured tool such as MCP lets Claude reason over candidate data in Leonar's own database while you keep every LinkedIn action in your own hands. ### Does Leonar replace LinkedIn Recruiter? No. Leonar is a separate workspace with its own database. Claude searches and ranks candidates there, contact data gets enriched, and multichannel outreach runs from one place. Your team keeps its Recruiter workflow inside Recruiter. Not sure which plan fits your team? Answer 3 quick questions and we will recommend the best option for your hiring workflow. How large is your recruiting team? What is your main challenge? Are you an agency or an in-house team? Your recommended plan [See pricing details](https://www.leonar.app/pricing/) sourcing linkedin-recruiter ai claude ![Pierre-Alexis Ardon](https://www.leonar.app/_astro/pierre-alexis-ardon.ByBiMn-t_2wY0oT.webp) Author [Pierre-Alexis Ardon](https://www.leonar.app/authors/pierre-alexis-ardon/) Co-founder Pierre-Alexis Ardon is co-founder of Leonar, where he focuses on building AI-powered recruiting systems, assisted sourcing, and search optimization. With a background in engineering and over 7 years working at the intersection of artificial intelligence and talent acquisition, he designs the algorithms that power Leonar's candidate matching and workflow assistance. Pierre-Alexis advises recruitment agencies on their digital transformation and regularly publishes analyses on how AI agents are reshaping HR workflows. He is passionate about making advanced technology accessible to recruiters who are not engineers. AI recruiting systems Sourcing automation Recruiting analytics AI agents for HR [LinkedIn](https://www.linkedin.com/in/ardonpa/) ## Related articles - ![7 Sourcing Techniques on LinkedIn Recruiter](https://www.leonar.app/_astro/7-sourcing-techniques-linkedin.Dn1kBYWH_Zp2Uhs.webp) Sourcing ### [LinkedIn Recruiter Best Practices: 7 Sourcing Techniques](https://www.leonar.app/blog/7-sourcing-techniques-linkedin/) - ![Recruiter using Google X-Ray search to find public candidate profiles across GitHub and portfolio sites](https://www.leonar.app/_astro/google-x-ray-search.naF0SHiM_GHVMv.webp) Sourcing ### [Google X-Ray Search: The Recruiter's 2026 Guide](https://www.leonar.app/blog/google-x-ray-search/) - ![Sourcing vs recruiting differences for recruiters and hiring teams](https://www.leonar.app/_astro/sourcing-vs-recruiting.CxScdTHF_ZoQD2g.webp) Sourcing ### [Sourcing vs Recruiting: Differences and Why It Matters](https://www.leonar.app/blog/sourcing-vs-recruiting/) ## Streamline your recruitment with AI & automation Join 400+ recruiting teams already using Leonar to source, engage and hire faster. 7-day free trial with full access 7-day free trial Full feature access Cancel anytime