Most guides about AI and job searching are theoretical. "AI will change how we find jobs." "The future of hiring is conversational." Lots of hand-waving, not a lot of showing.
This is the opposite. This is what it actually looks like to run a job search through a single conversation with Claude, connected to Remoet via MCP (Model Context Protocol). I'm going to walk you through each step with real interactions you can replicate today. Think of it as a single-conversation workflow: profile, discovery, shortlist curation, and keeping every role on file, all without leaving your AI agent.
If you haven't set up MCP yet, our setup guide for Claude, Cursor, VS Code and Codex will get you connected in about five minutes.
Why This Is Different From AIApply, Jobright, and Auto-Apply Tools
I want to address the elephant in the room before we start. There's a whole category of AI job search tools that promise to auto-apply to hundreds of jobs for you. Jobright, AIApply, LazyApply, and others. They optimize for volume.
The problem with volume-based auto-apply is that it optimizes the wrong half of the process. Spraying 500 generic applications gets you filtered out by the same ATS systems these tools claim to beat, and it burns the goodwill of every company you would actually want to work for.
What we're doing here is a different shape entirely. Remoet is an MCP server for job search, which means your AI agent has real tools to search the open board, research companies, evaluate fit, build a shortlist, and keep what it found on file. It's conversational job search: you talk to Claude, Claude uses Remoet's MCP tools, and together you build a targeted pipeline. Companies on Remoet carry a tech stack that is scraped and enriched from the roles they are actually hiring for, not keywords a recruiter pasted onto a posting. When you search for "TypeScript and Node.js," you're matching against that, not against job descriptions that happen to contain the word "Node."
Let me show you how it works.
Step 1: Connect and Check Your Profile
The first thing to do after connecting is see where you stand. Open Claude and say:
"Show me my profile."
The agent pulls your complete developer profile from Remoet: your summary, skills, work experience, education, projects, and any gaps. If you're starting fresh, most fields will be empty.
Here's what Claude might come back with:
"Your profile is mostly empty. You have no summary, no skills listed, no work experience entries, and no projects. Your profile won't be competitive without at least a summary, skills, and some work history."
This isn't a canned response. The agent is reading your actual data and evaluating what's missing. On a traditional job board, you'd have to notice these gaps yourself by staring at blank form fields. Claude tells you upfront.
Step 2: Build Out Your Profile Through Conversation
Now you fill in the gaps, but instead of tabbing through form fields, you just talk:
"Update my summary to say I'm a senior full-stack developer with 6 years of experience building web applications with TypeScript, React, Node.js, and PostgreSQL. I'm based in Berlin and open to remote or hybrid roles in Europe."
The agent updates your summary. Next:
"Add a work experience entry. I worked at Acme Corp from January 2022 to March 2025 as a Senior Software Engineer. I led the migration of their monolithic PHP backend to Node.js microservices, reduced API response times by 40%, and mentored a team of 3 junior developers."
One conversational message instead of filling out 6 form fields. Keep going:
"Add TypeScript, React, Node.js, PostgreSQL, and Docker to my skills."
You can layer in education, projects, whatever you want. The point is that building a complete profile through conversation takes maybe 10 minutes. Through forms, it takes 30 to 45 minutes because you're context-switching between fields, formatting text, and fighting with date pickers. I've done it both ways. The conversation approach saves time, and it's a lot less draining too.
Step 3: Search the Whole Job Catalogue
Before you star anything, look at what's actually open. The Remoet job catalogue is public, so your agent can search all of it from a brand new account with zero stars:
"Search Remoet for platform engineering roles that use Go and Kubernetes."
That runs search_jobs, which reads the same open board anyone can browse at remoet.dev/jobs. Every company in the catalogue, no star needed and none consumed. Each result comes back with the role's public Remoet page (they live at /listings/<company-slug>/jobs/<role-slug>, real pages you can open or send to a friend), the employer's own application URL, the company slug, the tech stack, the salary if the posting states one, the remote policy and the seniority.
This is the fastest way to get something useful out of a fresh connection. You have results in one message, before you've made a single decision about which companies you like.
Three things are worth knowing before you lean on it. The filters are real and run on the server: search_jobs takes free text (matched against the title, Remoet's own summary of the role and the stack), a list of technologies, a match mode so you can demand every technology on the same role rather than any of them, a company slug, location, remote policy, experience level and a minimum salary, plus sorting by newest or salary. So this is one question, not a scan:
"Senior Go roles in Germany or the Netherlands, remote or hybrid, above 90k, newest first."
Location matches the canonical places Remoet stores per role, so asking for Germany also finds the roles that only name Cologne or Berlin. A salary floor drops every role with no published salary, which is roughly 45% of our board, so use it when salary is a hard requirement and leave it off otherwise.
When nothing matches, the answer names the filter that emptied the search and says how many roles you would get back by dropping it, so you know what to loosen instead of guessing. And the coverage is honest, not universal: the public catalogue is scraped from cleared ATS platforms, Lever, Greenhouse, Ashby, Recruitee and Workable, so a company running its careers page on something else isn't in there yet, no matter how much you'd like to work for them. Duplicate postings of the same role at one company get collapsed into a single result, so you're reading roles, not postings.
Step 4: Find Companies That Match Your Stack
Searching jobs tells you what's open this week. Searching companies tells you where you'd want to be for the next three years. Different question, different tool.
"Find companies that use TypeScript and Node.js."
The agent searches Remoet's company catalogue and returns results ranked by relevance to your query. It's matching against the stack Remoet derives from the roles each company is hiring for, not keyword-scraping job descriptions. That's a meaningful difference: it reflects what a company is staffing right now.
You'll get a list like:
"Here are the closest matches for TypeScript and Node.js: [Company A] (TypeScript, Node.js, React, 12 open roles), [Company B] (TypeScript, Node.js, Go, 8 open roles), [Company C] (TypeScript, Node.js, AWS, 5 open roles)..."
Three technologies each, because that is what an unstarred listing previews. Star a company and you see its whole stack.
Already better than scrolling through 200 generic listings. But you can refine further:
"Which of those also use PostgreSQL?"
"Show me more details about Company A. What do they build?"
Claude pulls the listing profile for any company you're curious about: their description, the previewed stack, perks, URLs, and how many open jobs they have right now. This is the conversational part. You're not clicking through pages. You're having a back-and-forth that narrows down your options naturally.
Step 5: Star the Companies You Care About
Once you've found companies that genuinely match your background, star them:
"Star Company A and Company C."
Worth being precise about what starring actually does, since the word invites the wrong picture. It isn't a lock on the jobs: search_jobs already reaches every role at every company, no star required, filters and all. Starring is closer to a subscription: the company's new postings start showing up in your feed and the roundup email, and the stack view stops truncating at three technologies and shows you everything. Think standing order, not bookmark.
Which is exactly why you should be selective. A star you didn't mean costs you noise every week, not nothing.
Claude knows this. If you try to star a company whose stack doesn't overlap with your skills, it'll flag it:
"Company X primarily uses Rust and Go. Your profile lists TypeScript and Node.js with no Rust or Go experience. Starring this company would add their jobs to your feed, but you may not be qualified for most of them. Want to star it anyway?"
That kind of judgment call is something a traditional UI simply can't make. The agent is cross-referencing your profile with the company's tech stack in real time, and it asks before it spends anything on your account.
You get 50 active stars, which is room for a real shortlist. Unstarring is budgeted at 25 per 30 days, so churning through the catalogue and re-starring on a whim isn't the intended move.
Step 6: Check Your Job Feed
After starring companies, you've got a personalized feed. Check it:
"Show me jobs from my starred companies."
This runs get_starred_jobs: the same filtering as the open board, scoped to the companies you chose, so you can go straight to the specific question.
"Only senior-level roles, minimum 90k."
"Any of these that are remote with no location restrictions?"
You're having a conversation with your job feed instead of scrolling through it. The difference sounds small, but it changes how you engage: you actually read and evaluate instead of skim and forget.
If the feed comes back empty, that's not a bug. It means you have no stars yet, or the companies you starred aren't hiring right now. Go back to Step 3 and search the open catalogue directly.
Step 7: Save Jobs and Take Notes
When something catches your eye, save it. This works on anything, whether it came from the public search or your feed:
"Save the Senior Full-Stack Engineer role at Company A."
Then add context so future-you remembers why:
"Add a note: their stack is a near-perfect match, they have an engineering blog I like, and the role mentions working on their new API platform which is exactly what I did at Acme Corp."
This is your private CRM. Three weeks from now, when you're comparing five companies and trying to remember which one excited you and why, these notes are invaluable. And because you're adding them in the moment through conversation, the friction is zero compared to maintaining a spreadsheet. I've tried the spreadsheet approach. It lasts about four days before you stop updating it.
Step 8: Get the Application URL and Apply on the Company Site
This is the step most "AI job search" content gets wrong, so I want to be explicit about it.
Almost every job on Remoet is scraped from an external careers page. When you tell your agent you want to apply, the agent's job is to hand you the application URL, not to submit anything on your behalf. You click through, you fill out the company's form on their site, you hit submit. The application happens in the company's system, not in Remoet.
"I'm ready to apply to the Senior Full-Stack Engineer role at Company A. Give me the application URL and anything I should know before I go."
The agent surfaces the URL, the role description, the company's tech stack, and any notes you've already attached. If your profile has gaps that would be obvious on a CV, the agent flags them:
"Your profile has no projects listed. Companies in this stack often look for portfolio links. Want to add a project before you apply?"
You decide. Then you go apply on the company site like you always have.
The exception: if a company posts a role directly on Remoet through the partner system, the agent can submit end to end from your profile, and it will tell you when a role is that kind:
"This role is an internal job. I can submit your application through Remoet using your profile, or you can apply via the company URL. Which do you want?"
Nearly every role on the board is an external scrape, so in practice the answer is "I'll apply on the company site." That's not a failure of the agent, it's the honest shape of an internet where most jobs live on external ATSs. What the agent gives you is everything around the apply click: discovery, evaluation, notes, memory. Application submission, for now, is still on you.
Step 9: Keep Every Role On File
Once you've applied, put that on the record where your agent will see it next time:
"I applied to the Senior Full-Stack Engineer role at Company A today through their site. Save it and note the date and that I mentioned the API platform work."
Saved jobs are your agent's memory across sessions, and the note field is where the state of play goes. Ask for it back any time:
"Show me my saved jobs and their notes."
"Which of my saved roles did I apply to, and when?"
Then keep the notes current as things move:
"Update the note on the Company A role: had a first-round call, the interviewer mentioned they're expanding the platform team."
Everything stays attached to the saved role. When you check back next week, all the context is there. No more digging through email threads trying to reconstruct timelines.
A note on scope, because I'd rather be straight with you: Remoet does have full application records with status timelines, and they work on the website for internal jobs. They are not exposed to agents yet. Saved jobs plus notes is the agent-side pipeline today.
Step 10: Weekly Check-Ins
The real power of this workflow isn't the initial setup. It's the ongoing maintenance. Once your agent is connected, check-ins take about five minutes:
"Anything new from my starred companies this week?"
This is the payoff for starring. The agent checks your feed and the latest roundup email (which arrives daily, weekly on Mondays or monthly, whichever cadence you picked) and shows what landed since last time, without you going looking. You scan the list, save anything interesting, and move on. Compare that to the old workflow: open 5 job board tabs, search each one, scroll through results, try to remember what you already saw last week.
And once in a while, go back out to the open catalogue for the things your stars would never surface:
"Search the whole catalogue for staff-level backend roles that mention Postgres, remote, and skip anything I've already saved."
"Remind me what I've got saved and what I noted about each one."
Quick status check across your pipeline. No dashboard required.
The Full Loop
Here's the complete workflow:
Connect your AI agent to Remoet via MCP (one-time setup)
Audit your profile and fill in gaps through conversation
Search the public job catalogue for what's open right now, no stars needed, with the filters that matter
Search for companies by tech stack, for where you'd want to be long term
Star the companies that match your background, so their new roles come to you
Check your feed for what those companies posted since last time
Save interesting roles with notes
Get the application URL and apply on the company's site
Note where you applied and what happened, on the saved role, so it stays on file
Repeat steps 6 through 9 on your own cadence, with an occasional sweep of the open catalogue
Total time: about half an hour to get from zero to a curated pipeline with your first application sent. After that, maybe 5 minutes a week to stay on top of everything.
On a traditional job board, the initial setup takes hours, and the upkeep takes 30 to 60 minutes of browsing, filtering, and clicking every time you sit down. The agent compresses that into conversation, and you still do the apply click on the company site for most roles.
Questions people ask me about this
Do I have to star companies before I can see any jobs?
No. The job catalogue is public. search_jobs reads the open board at remoet.dev/jobs across every company, with no account at all, and starring costs you nothing to browse. It carries the full filter set too: text, stack, company, location, remote policy, seniority, minimum salary and sorting. A star is a subscription to a company's new roles, not a key to them.
Can I do all of this for free?
Yes. Remoet is free for job seekers, full stop. Every account gets the same limits: 50 active starred companies, 25 unstars per 30 days, 5,000 MCP requests per day (an abuse ceiling, not a meter you will feel), 5,000 API requests per day, 10 link trees, and unlimited access to your profile and saved jobs. There is no paid tier to upgrade to.
What if I want to use the website for some steps?
Everything's synced. Star a company through Claude and it shows up on the website. Save a job on the website and your agent can see it. Use whichever interface feels right for each step. Some people prefer the website for browsing job descriptions and the agent for searching and keeping notes. Whatever works for you.
Does the agent remember my previous conversations?
Claude has access to your Remoet data (profile, stars, saved jobs, notes) across all sessions. Your data persists regardless of which conversation you're in. What Claude doesn't remember is the conversation itself from a previous session, but since all the important state is stored in Remoet, that rarely matters. Your notes are your memory.
Can I use a different AI agent than Claude?
Mostly. Cursor, VS Code, Codex, Gemini CLI, Cline and Windsurf (now Devin Desktop) all connect to Remoet, and so do Hermes and OpenClaw through the skill. The one exception is ChatGPT: our server enforces an Origin allowlist covering the Claude and Remoet hosts, so a ChatGPT connector is refused before it ever gets to auth. The experience is otherwise similar across clients, though some handle multi-step tool calls more smoothly than others. I've found Claude to be the most reliable for complex workflows, but Cursor is great if you're already in it for coding and want to job search from the same window.
Which companies are in the catalogue?
Remoet scrapes cleared ATS platforms: Lever, Greenhouse, Ashby, Recruitee and Workable. If a company hosts its careers page somewhere else, its roles aren't in the catalogue yet, and your agent won't find them however you phrase the query. It's the honest limit of the coverage, and it's worth knowing before you conclude a company isn't hiring.
So the agent doesn't actually submit my applications?
For the vast majority of roles, no. Almost every job on Remoet is scraped from a company's external careers page, and the actual application form lives on that company's site. The agent hands you the URL and the context. You click submit on the company's site. That is the honest shape of the workflow and any tool that claims otherwise is either lying or is restricted to a tiny in-network slice. If a company posts directly through Remoet's partner system, the agent can submit end to end using your profile, and it will tell you when that option exists. Either way, save the role and note what you did, so the pipeline stays in one place.
Now go run it
Copy the prompts above into your own agent and see how far you get in one sitting. Then tell me where it got awkward: a filter that didn't do what you expected, a step that needed three tries, a question I should have answered here and didn't. Discord is the fastest way to reach me, and the workflow above has changed more than once because someone walked it and said "this bit is weird."
