The first thing everyone built when agents got the ability to call tools was a firehose. Plug your agent into eight job boards, let it pull thousands of listings at machine speed, hand the pile to you. The pitch was "AI-powered job search." What you actually got was the same problem you already had, except now it arrives faster and there is more of it.

I run Remoet, which is one of the things on this list, so read accordingly. But I have connected my own agent to most of these, and the honest version of the landscape is more useful to you than another vendor ranking itself first. The existing comparisons out there are mostly written by data companies, and they all happen to conclude that the data company won. So here is the map without that particular thumb on the scale.

The firehose: JobSpy

JobSpy is the open-source one, and it is genuinely good at what it does. It scrapes across the big boards (LinkedIn, Indeed, Glassdoor, ZipRecruiter, Google Jobs, and a few regional ones), filters by title, location, posting date, remote status, and salary, and hands your agent structured results. Free. No account. Run it locally and point Claude at it.

If you want everything, this is everything. That is also the catch. JobSpy does not have an opinion about which of those thousands of jobs is worth your attention, because it cannot. It is a pipe. The judgment is still entirely yours, and "read every listing on the internet, but through an agent" is not actually a smaller job than it was before. For a lot of people that is fine. If you are casting the widest possible net and you trust yourself to sort it, start here.

The one that logs in as you: LinkedIn MCP Server

LinkedIn MCP Server is the popular open-source answer to "most of the jobs are on LinkedIn anyway." It drives a real browser session, either by importing cookies from your logged-in browser or walking you through a fresh login, and gives your agent profile lookups, company pages, job search, and job details. Free, Apache-licensed, actively maintained.

The caveat is the author's own, stated right in the README: LinkedIn's user agreement prohibits automated access, and accounts using automated tools can be restricted or banned. This is you automating your own account against the platform's wishes, with no guarantee of safety. Plenty of people accept that trade with their eyes open. Just know it is the trade, because no marketing page will tell you.

The market-data tools: TheirStack, Coresignal, Bright Data

This is the group people connect by accident and then wonder why it feels off. TheirStack, Coresignal, and Bright Data all expose enormous job and company datasets over MCP. Hundreds of millions of postings, technographic signals, company enrichment, the works. Pricing lands around fifty to sixty dollars a month once you are past the free tier.

Here is the thing nobody says plainly: these are not job-seeker tools. They are built for salespeople, recruiters, and analysts who want to know who is hiring, for what, using which technologies, in order to sell something or source candidates. The data is real and the MCP integrations are solid. But if you are a person looking for your next role, you are holding a market-intelligence instrument by the wrong end. Great tools. Wrong job.

The board with a front door: Dice

Dice shipped an MCP server over its tech-only job database, and it is a clean piece of work. Natural-language queries, a real set of filters (location radius, workplace type, employment type, visa sponsorship, posting recency), plus resume comparison and upskilling suggestions. It connects to Claude, ChatGPT, and Gemini.

If what you want is a tech job board that your agent can search instead of a website you click through, Dice is the most straightforward answer on this list. The mental model is still "search a board," just with a better interface. That is not a criticism. For a lot of people a good board they can talk to is exactly the upgrade they wanted.

The ATS aggregator: Career Site Jobs

Career Site Jobs (Fantastic.jobs) pulls postings straight from company career pages across a large set of ATS platforms like Workday, Greenhouse, and Ashby, then enriches them. The angle is freshness and going direct to the source rather than a board's copy of it. If your frustration is stale or duplicated listings, this is the one built for that specific complaint.

The self-hosted corner: JobSync and Job Hunter MCP

If the part you hate is keeping track of where you applied rather than finding the openings, the open-source world has that covered too. JobSync is the substantial one: a self-hosted tracker (one docker compose up) with resume management, job matching, and a built-in MCP server, so your agent can file applications and interview questions straight from the chat, with your approval on every write. Your data stays on your machine, and it can run against a local model through Ollama if you want the whole loop private. There are also smaller solo builds like Job Hunter MCP, which lean into fit scoring and lightweight tracking. Worth a look if you want something minimal.

A note on the open source ones

You may have noticed the pattern: the open-source tools on this list run on your machine. That changes what "MCP server" even means. When the data lives on someone else's server, MCP is the access. When the code runs locally, MCP is just one door into it, and some of the biggest open-source projects in this space skip the protocol entirely. The loudest example is AIHawk, an auto-apply bot with more than thirty thousand GitHub stars that logs into LinkedIn and fires tailored applications at Easy Apply listings. Not an MCP server, does not need to be, and whether it is a good idea is a question I get to further down.

So if you are comfortable running things locally, do not filter your search by "MCP server." Filter by what the tool actually does, then check whether your agent can drive it. Increasingly it can, protocol or not.

Where Remoet sits, and where it does not

Remoet is the opposite of the firehose, on purpose. It does not try to show you every job. The agent matches companies by their actual tech stack (the real stack we scrape and enrich at the company level, not the keyword soup recruiters paste onto a posting), you star the ones you would genuinely work for, and the job feed is scoped to that shortlist. Curated input, focused output. The star list is the whole point: it is a filter you build by talking, and it is the thing that keeps your agent's attention (and its tokens) on companies that actually fit.

The natural prompt is "I write Rails on top of React with Postgres, fully remote preferred, find my shortlist," and the agent does the rest: builds the list, scopes the jobs, manages your profile and applications, all in conversation.

Now the honest part, because the vendor comparisons skip this. Remoet does not apply to jobs for you. Most listings are scraped from companies' external career pages, so when you find one, you apply on their site like always. The free tier has caps on how many companies you can star. And if your profile is empty, stack matching has nothing to match against, so there is setup before there is magic. If you want a tool that auto-blasts applications, this is not it, and I would gently suggest that auto-blasting applications is part of why everyone's inbox is broken.

How to actually pick

Strip away the marketing and it comes down to what you are doing:

None of these is a scam and most of them are well built. The mistake is connecting the wrong category to your problem, which is easy to do when every one of them is filed under "job search MCP server" and half of them are not really for job seekers at all.

If the curation angle is the one that fits, you can connect Remoet to your agent in a few minutes. Grab a free key at remoet.dev/onboarding, point your client at the MCP server, and ask it to build your shortlist. If one of the others fits better, use that one. The good news, for the first time in a while in job search, is that there is now a real choice.