Clover Public
io.github.fourleafaiPublished by GitHub account fourleafaiv1.1.0Updated Oct 27, 2026
Job search and interview prep MCP. 15 tools, OAuth 2.1, cross-LLM. four-leaf.ai.
Overview
Hosted job-search and interview-prep MCP that gives an assistant live job postings, role intel, answer scoring, resume matching and comp research.
- What you can do
- Search 140,000+ active job postings in natural language and open a full posting
- Generate role-calibrated practice questions and score typed answers against a five-dimension rubric
- Score a resume against a job description and flag gaps
- Research compensation, including cited salary bands and offer analysis
- Read and write an application tracker
- Pull role-specific interview intelligence and strategy guidance
- When to use it
- Useful when someone is actively job hunting or preparing for a specific interview and wants the assistant to work from real postings, rubric-scored practice and live salary data instead of generic advice.
- Requirements
- Remote HTTP endpoint with OAuth 2.1; the first tool call opens a browser sign-in. A free Four-Leaf account is needed for the data tools (3-day trial, no credit card). The optional coaching Skill is installed separately via npx and works without the MCP in a reduced, coaching-only mode.
Installation
In SourceWeft
- Open Clover Public in the dashboard and add it to a workspace.
- Connect the required credentials. They are stored privately and never shown on public pages.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Use with an agent
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.fourleafai%2Fclover-public/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Other MCP clients
Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.
{
"mcpServers": {
"clover-public": {
"type": "http",
"url": "https://four-leaf.ai/api/mcp"
}
}
}README
four-leaf-coach
An open-source Skill that turns Claude (or ChatGPT, Cursor, Codex, GitHub Copilot) into a job search and interview prep coach. It pulls real job postings, role-specific interview intelligence, rubric-scored feedback on practice answers, and resume scoring from the hosted Four-Leaf MCP, then walks the user through preparing for an actual interview.
Free to install and use, including scoring the answers you type. Voice mock interviews (spoken answers, adaptive follow-ups) and full AI resume tailoring live on four-leaf.ai; the Skill surfaces those as an upgrade path when they're the right next step.
What it does
Walks a user through prep for a specific role at a specific company. The Skill greets, asks what they're prepping for, and routes them into one of seven guided workflows. Every workflow pulls live data from the Four-Leaf MCP's fifteen tools (jobs and full postings, role intel, question bank, answer scoring, match scoring, comp research, application tracker) and adds the Skill's coaching on top.
You don't type these as commands. Say what you want in plain language and the coach routes to the right workflow:
- Kickoff the coach reads your application tracker, figures out what you're prepping for, and routes you. Say "help me get ready for my job search" or "where am I with everything".
- Find jobs natural-language search across 140,000+ active postings, then pull up a full posting and save it to your tracker. Say "find me remote senior data scientist roles".
- Prep for a role interview pipeline, what to expect, how to win. Say "what's the interview like for a PM at Stripe".
- Practice calibrated questions, then type your answer and get it scored 1-10 against the same five-dimension rubric the Four-Leaf practice page uses. Say "give me a few hard system design questions".
- Analyze a JD scores a resume against a posting and points out gaps. Say "how does my resume stack up against this JD".
- Negotiate and comp research real numbers from the MCP. Have an offer? It runs a full analysis (total comp, market percentile, red flags, a counter strategy with exact talking points). Just asking what a role pays? It returns a cited salary band from a live web search. Say "they made an offer, help me negotiate", "is this offer any good", or "what's a good salary for a backend engineer in Austin".
- Interview strategy formats, AI interviewers, work trials, signal vs noise. Say "what are AI interviews actually like".
Install
Two steps. The Skill tells your AI tool how to coach. The hosted MCP gives the Skill live data.
Step 1: install the Skill
That's it. The CLI detects your tool (Claude Code, Cursor, Codex, or GitHub Copilot), asks for scope when it matters, and copies the right bundle into the right place. Useful flags:
--tool <name>pick the tool yourself:claude-code,cursor,codex, orgithub-copilot--scope <project|global>Claude Code only; where to install--dry-runshow what it would do without writing anything--yesskip confirmation prompts,--forceoverwrite an existing installfour-leaf-coach listshow the supported tools and what's detected in the current directory
Cursor needs the Nightly channel with Settings, Rules, Agent Skills enabled. The other tools work out of the box.
Step 2: install the Four-Leaf MCP for live data
The Skill works in degraded mode (coaching only, no live job data) without the MCP. To get real job search, role intel, answer scoring, and resume scoring, install the hosted MCP:
The first tool call opens the browser for OAuth. A free Four-Leaf account works (3-day trial included, no credit card).
For Cursor, ChatGPT Desktop, and other MCP-aware tools, configure the same URL (https://four-leaf.ai/api/mcp) per that tool's MCP setup docs.
Step 3: use it
Once it's installed, just tell the coach what you need in plain language. Say something like "help me prep for a senior software engineer interview at Google" and it takes it from there. There are no subcommands to memorize; the coach routes based on what you ask. (In Claude Code you can also invoke it explicitly with /four-leaf-coach.)
Manual install
If you'd rather not run npx (air-gapped network, or you just want to see what lands where), clone and build the bundles yourself. This is exactly what npx four-leaf-coach add --tool <name> automates.
npm run build reads skills/four-leaf-coach/SKILL.md plus the adjacent references/ tree and writes a dist/<tool>/ directory for each supported tool. Then copy your tool's bundle into place:
The GitHub Copilot bundle is a single flattened file (.github/copilot-instructions.md) because Copilot reads one instructions file and doesn't follow references. The build inlines the whole Skill for it. The other three tools follow file references, so they get the source tree as-is.
What's free vs paid
- Free: the Skill itself, and thirteen of the fifteen MCP tools. That's jobs and full postings, role intel, the question bank, answer scoring, match scoring, comp research, and reading and writing your application tracker. Six of them carry a daily cap on a free account (job search 30/day; question generation, answer scoring, match scoring, and the two comp tools 20/day each, on separate counters). The rest are unmetered.
- Paid on four-leaf.ai: two tools. Voice mock interviews with spoken answers and adaptive AI follow-ups, and full AI resume tailoring against a specific JD. A paid plan also lifts the daily caps above. Three options at four-leaf.ai/pricing. 3-day free trial (no card), $5 5-Day Pass, $20/mo Pro. All three give you the same features.
The Skill surfaces these when they're the right next step. It doesn't push.
FAQ
What is four-leaf-coach? An open-source Skill that turns Claude, Cursor, OpenAI Codex, or GitHub Copilot into a job-search and interview-prep coach. The Skill is a structured set of instructions plus reference files that your AI tool loads. It works with the hosted Four-Leaf MCP server for live data (real job postings, role-specific interview intelligence, answer scoring, resume scoring, your application tracker).
How is this different from just prompting Claude for interview prep?
Generic prompts produce generic advice. four-leaf-coach calls real tools: search_jobs returns actual apply URLs from 140,000+ active postings and get_job expands one to the full text; match_score runs a real scoring algorithm against your resume; generate_practice_questions produces role-calibrated questions and score_answer grades what you type against a five-dimension rubric; save_application puts a role in your tracker. The Skill orchestrates the calls and coaches around them.
Do I need a Four-Leaf account? For the data tools (jobs, role intel, question bank, answer scoring, match scoring, comp research, application tracker) a free Four-Leaf account works. The 3-day trial requires no credit card. Voice mock interviews and full AI resume tailoring are paid features on four-leaf.ai. The Skill surfaces those as an upgrade path when relevant.
Does the MCP work with ChatGPT? Yes. The MCP is HTTP-based with OAuth, so any MCP-aware client connects: Claude Desktop, Claude Code, Cursor, ChatGPT Desktop (Plus + Dev mode), Cline, Continue, Windsurf. The Skill itself is a Claude / Cursor / Codex / Copilot primitive; ChatGPT doesn't yet have a comparable file-based Skill convention, but the underlying MCP tools work there.
Can I use the Skill without the MCP? Yes, in degraded mode. Without the MCP, the Skill still coaches with structured workflows for prep, practice, JD analysis, and negotiation, just without live job listings, rubric-scored answers, or real resume scoring. Install the MCP to unlock the live data path.
Is the question bank really open? The Skill, the per-tool dist pipeline, and the install CLI are MIT-licensed in this repo. The question bank itself lives behind the MCP today; the open-data play is on the roadmap.
What about Pi, Gemini CLI, OpenCode, Trae, Rovo Dev, Qoder? On the roadmap. The dist pipeline can target any AI tool with a documented Skill or instructions-file convention. PRs welcome.
Repo layout
skills/four-leaf-coach/ is the source of truth. dist/ is generated output, so edit the source and rerun npm run build.
Roadmap
- Registry submissions to every Skill aggregator that ships a public registry.
- Coverage for more tools as their Skill conventions stabilize: Pi, Gemini CLI, OpenCode, Trae, Rovo Dev, Qoder.
Done so far: per-tool dist/ bundles generated from a single source (npm run build), a flattened single-file variant for GitHub Copilot, and the npx four-leaf-coach add one-command installer.
Contributing
PRs welcome that improve a command's coaching, add a new command, or fix a voice issue. Anything that drifts the Skill's positioning away from "coach, not cheat tool" will be declined.
License
MIT. See LICENSE.
Source: README.md at commit af34293
Tools
0Version history
2- v1.1.0LatestSep 22, 2026
- v1.0.0Sep 16, 2026
Score
85% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
- Auth declaredRequired10 pts
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts