Your AI job-search agent — search everywhere, apply and track in one place.
career-agent turns any MCP-capable chat — Claude Desktop, Claude Code, ChatGPT, Cursor, Gemini CLI, and others — into a personal job-search assistant. Ask in plain language, in any language, and it will search live job listings across dozens of job boards and thousands of company career sites. For jobs on the Flowxtra platform it goes one step further than every other job-search tool: sign in with your Flowxtra candidate account and apply, track your applications, and manage your profile — without leaving the chat.
Search anywhere. Apply where it counts. Most AI job tools stop at "here's a link." career-agent closes the loop on Flowxtra jobs — one confirmed message and you've applied, with the status tracked from the same conversation.
- Why career-agent
- What it can do
- Install
- Add the MCP server to your tool
- Sign in to apply and track (optional)
- Example prompts
- Sources
- How it works
- Privacy & safety
- Attribution
- Contributing
- Disclaimer
- License
| Typical AI job search tools | career-agent | |
|---|---|---|
| Search many boards & ATS platforms | ✅ | ✅ |
| Read full job descriptions | ✅ | ✅ |
| Apply to a job | ❌ (link only) | ✅ on Flowxtra (with your confirmation) |
| Track application status | ❌ | ✅ on Flowxtra |
| Works from your existing chat client | ✅ | ✅ |
| No API keys, no scraped data | ✅ | ✅ |
career-agent is read-only by default — searching needs no account. The apply-and-track features unlock when you sign in with a free Flowxtra candidate account, and applying always requires your explicit confirmation for each specific job. The agent prepares; you approve. Never a spam bot.
Search & read (no account needed)
- Search live listings across job boards, big-tech careers sites, and 3,500+ companies on 17 ATS platforms — all behind a few tools.
- Read a full posting: description, location, salary, employment type.
Apply & track on Flowxtra (sign in)
flowxtra_apply_to_job— submit your application to a Flowxtra job using the CV on your profile. Requires an explicit per-job confirmation.flowxtra_my_applications— see all your Flowxtra applications and each one's current pipeline stage, across every company.flowxtra_my_profile— check your profile basics and whether a CV is on file.
Homebrew
brew install --cask Flowxtra/tap/career-agent
Go
go install github.com/Flowxtra/career-agent/cmd/career-agent@latest
Upgrade by rerunning the same command; pin a version with @vX.Y.Z.
Prebuilt binaries — download the archive for your platform from
Releases and put
career-agent on your PATH.
Docker (multi-arch: linux/amd64, linux/arm64)
docker pull ghcr.io/Flowxtra/career-agent
With career-agent on your PATH:
Claude Code
claude mcp add career-agent -- career-agent
Codex
codex mcp add career-agent -- career-agent
Gemini CLI
gemini mcp add career-agent career-agent
For Claude Desktop, ChatGPT, Cursor, and other GUI clients, add a server
entry that runs the career-agent command over stdio (see your client's MCP
settings).
With Docker instead, replace career-agent with
docker run -i --rm ghcr.io/Flowxtra/career-agent:
claude mcp add career-agent -- docker run -i --rm ghcr.io/Flowxtra/career-agent
💡 Searching multiple boards and fetching full descriptions can use a lot of tokens. A cheaper/faster model is fine for routine searching.
Searching works immediately. To apply and track applications, sign in once with your free Flowxtra candidate account:
career-agent auth login
You'll be prompted for your email and password (the password is never echoed
and never stored — only the resulting session token is saved, 0600, in your
user config directory). Then restart your MCP client so it picks up the
session.
career-agent auth status # check whether you're signed in
career-agent auth logout # remove the stored token
Prefer not to store a token? Set FLOWXTRA_TOKEN in the server's environment
instead.
Once the server is added, just talk to your assistant:
- "Find remote sales jobs in Europe."
- "Show me backend engineering roles at companies using Greenhouse."
- "What does this posting require?" (paste a job URL)
- "Apply me to this Flowxtra job." → the agent shows the job and asks you to confirm before submitting.
- "What's the status of my applications?"
- Job boards: 104, Cake, Jobindex (Denmark), マイナビ転職 (Japan), Flowxtra (board-wide across every company on the Flowxtra ATS), LinkedIn, and Indeed (public search).
- Company career sites: 3,500+ companies on Workday, Ashby, Greenhouse, Lever, Teamtailor, Recruitee, Eightfold, SAP SuccessFactors, SmartRecruiters, Workable, Rippling, iCIMS, Oracle Recruiting Cloud, JOIN, UKG Pro (UltiPro), Avature, and BambooHR — all behind one company-search tool. A company outside the built-in roster works too: pass its careers-page URL (except Eightfold, SAP SuccessFactors, and JOIN, which support roster companies only).
- Dedicated big-tech sites: Amazon Jobs, Apple Careers, Google Careers, Meta Careers, NVIDIA Careers, and TSMC Careers.
career-agent is a single Model Context Protocol server that exposes job-search and candidate tools over stdio. Your chat client launches it, the model decides which tools to call, and results come back as structured data the model presents to you. Every request goes out from your own machine directly to each site's public endpoints — there is no career-agent backend in the middle, and nothing is scraped, cached, or redistributed.
The candidate tools call Flowxtra's public candidate API with your session token, exactly as the Flowxtra website would.
- Your machine, your requests. No proxy, no central server, no telemetry.
- No scraped data. Listings are fetched live when you invoke a tool and stored nowhere.
- Credentials stay local.
career-agent auth loginsaves only a session token (0600), never your password. - Applying is always confirmed. The apply tool refuses to run without an explicit, per-job confirmation — the model cannot apply on its own initiative.
- Read-only elsewhere. career-agent only ever reads from third-party boards; it never logs in to them, so your accounts on those sites are never touched.
Use it at a human pace. Some sites may throttle an IP that sends too many requests too quickly; the server does not rate-limit on your behalf.
career-agent's multi-source search engine began as a fork of
amikai/openings-mcp (MIT), and the
full upstream commit history is preserved in this repository. Enormous thanks to
that project and its contributors — the provider architecture, the ATS roster,
and the search tooling are their work. The candidate-account layer (apply,
track, profile) and the career-agent auth CLI are first-party
Flowxtra additions built on top.
openings-mcp in turn credits these MCP job-search servers and data sources, carried forward here:
- job104-mcp, jobspy-mcp-server, mcp-jobs, 104-mcp-server
- OpenPostings — company/ATS discovery data used to expand the curated roster
- ai-job-search — Jobindex search reference
Issues and pull requests are welcome. New job-source providers follow the
pattern in internal/provider/ and internal/openingsmcp/; run go test ./...
before opening a PR. For Flowxtra candidate-facing changes, see
internal/provider/flowxtracand.
This tool is not affiliated with, endorsed by, or sponsored by any of the third-party job boards or companies whose public listings it searches (104, TSMC, LinkedIn, Indeed, and others). It is an independent job-search client for personal use. The Flowxtra integration is first-party.
Some providers rely on undocumented public APIs that may change or stop working at any time. Listing data belongs to the respective sites; the server fetches it on your behalf and redistributes nothing. You are responsible for complying with the terms of service of each site you query — do not use this tool for bulk scraping or data harvesting.