Skilark — MCP Server

Skilark MCP Server adds real-time job market intelligence to your AI assistant. It gives Claude, ChatGPT, and Gemini awareness of skill demand, hiring trends, job opportunities, and mock interview prep — context they don't have on their own.

What You Can Ask

What are the hottest skills right now?
Show me senior Rust jobs at Stripe
Watch for ML platform roles at Anthropic and alert me when new ones appear
Book me a behavioral mock interview for tomorrow at 2pm
Any interesting hiring signals this week?
I'm a Python backend engineer thinking about ML. What should I learn?
Show my interview feedback from last week
Any new jobs matching my saved searches?

Setup

Claude

Go to Settings → Connectors → Add custom connector.

Set the name to Skilark and paste the URL below. Leave OAuth fields blank.

https://mcp.skilark.com/mcp

On first connect, you'll be prompted to sign in with GitHub, Google, or email magic link.

ChatGPT

Go to Settings → Apps → Advanced settings → Create app.

Set the name to Skilark and paste the MCP Server URL below. You'll be prompted to sign in with GitHub, Google, or email magic link.

https://mcp.skilark.com/mcp

This creates a dev mode app. To use it in a chat, type /skilark to invoke the tools.

Gemini CLI

Add to ~/.gemini/settings.json (global) or .gemini/settings.json in your project root (project-level overrides global):

{
  "mcpServers": {
    "skilark": {
      "type": "http",
      "url": "https://mcp.skilark.com/mcp"
    }
  }
}

Then authenticate by running /mcp auth skilark in the CLI. Follow the browser link to sign in with GitHub, Google, or email magic link. Tokens are stored in ~/.gemini/mcp-oauth-tokens.json.

Note: gemini mcp add defaults to stdio transport — use the manual config above instead.

Available Tools (14)

`get_skill_trends`

Get the most in-demand tech skills ranked by job listing count.

Parameter	Type	Description
`days`	int, optional	Only count listings posted within this many days.
`limit`	int, default 20, max 50	Maximum number of skills to return.

`get_company_trends`

Get hiring activity across major tech companies, ranked by total open listings.

Parameter	Type	Description
`days`	int, optional	Only count listings posted within this many days.

`browse_jobs`

Search job listings with structured filters. Returns matching listings plus aggregated stats — top skills, seniority breakdown, and salary range. No raw descriptions: titles, companies, links, and analysis only.

Parameter	Type	Description
`skills`	list, optional	Filter by required skills. Must match canonical names: Python, Rust, Go, Java, TypeScript, JavaScript, C++, Scala, SQL, PyTorch, TensorFlow, JAX, Transformers, LLMs, NLP, Computer Vision, MLOps, RAG, Kubernetes, Docker, Terraform, AWS, GCP, Azure, PostgreSQL, Redis, Kafka, Spark, Airflow, Snowflake, Databricks, dbt, Flink, React, GraphQL, Spring Boot, Linux, Distributed Systems, System Design, CI/CD, Git, REST APIs, Microservices, gRPC, Elasticsearch, MongoDB, DynamoDB, Cassandra, BigQuery, Tableau, Looker.
`seniority`	string, optional	Filter by level: intern, junior, mid, senior, staff, principal, director, vp, c_level.
`role_category`	string, optional	Filter by role type: swe, ai_ml, data_eng, platform, devops, security, product, design, research, other.
`company`	string, optional	Filter by company slug, e.g. `stripe`, `openai`.
`location`	string, optional	Filter by location. Supports US states (`california`), cities (`san francisco`), and `remote`.
`days`	int, optional	Only include listings posted within this many days.
`limit`	int, default 20, max 50	Maximum number of listings to return.

`get_signals`

Market intelligence signals — hiring surges, salary insights, layoffs, funding, acquisitions, leadership changes, and more (12 signal types).

Parameter	Type	Description
`limit`	int, default 5	Number of signals to return. Pass 0 for all (max 100).

`get_nugget`

Get a bite-sized learning nugget from the tech refresher library.

Parameter	Type	Description
`category`	string, optional	Filter by topic area: languages, frameworks, data, devops, dsa.

`find_my_fit`

AI-powered semantic job search. Describe your ideal role in plain English. Slower than browse_jobs — uses vector search under the hood. Rate limited to 10 queries per minute.

Parameter	Type	Description
`query`	string, required	Natural language description of your ideal role.

`watch_query`

Save a job search to monitor — get notified when new matches appear. Runs daily with semantic matching. Up to 5 active watches per account.

Parameter	Type	Description
`label`	string, required	Human-readable name (e.g. "Senior Rust at Stripe").
`query`	string, required	Natural language description of the role to watch for.
`seniority`	string, optional	Filter by level: intern, junior, mid, senior, staff, principal, director, vp, c_level.
`company`	string, optional	Filter by company slug (e.g. `stripe`).
`location`	string, optional	Filter by location (e.g. `remote`, `california`).

`get_job_alert`

Check for new job matches from your saved watches. Returns listings found since the last check, grouped by watch name. Alerts are marked as seen after retrieval.

No parameters required.

`manage_watches`

List, pause, resume, or delete your saved job searches.

Parameter	Type	Description
`action`	string, required	One of: `list`, `delete`, `pause`, `resume`.
`watch_id`	string, optional	Required for delete/pause/resume. Get IDs from `action="list"`.

`book_interview`

Book a mock interview session. Standard types (behavioral, system design, etc.) require an invite code. quick_chat is a casual 10-minute intro with no invite needed (1/day limit). A Zoom meeting link is auto-created.

Parameter	Type	Description
`role_title`	string, required	The role to practice for (e.g. "Senior Software Engineer").
`interview_type`	string, required	One of: behavioral, technical_coding, system_design, product_sense, analytical, case_study, portfolio_review, role_play, quick_chat.
`scheduled_at`	string, required	RFC3339 UTC timestamp (e.g. `2026-04-08T14:00:00Z`). Must be in the future.
`invite_code`	string, optional	Required for all types except `quick_chat`.
`company_slug`	string, optional	Company to tailor the interview (e.g. `google`).

`get_interviews`

List your mock interviews — upcoming, in-progress, and completed. Shows status, type, and whether AI feedback is available.

No parameters required.

`get_interview_feedback`

Get detailed AI-generated feedback for a completed interview. Includes score, summary, strengths, improvements, and per-question breakdown.

Parameter	Type	Description
`interview_id`	string, required	The interview UUID (from `get_interviews` results).

`cancel_interview`

Cancel a scheduled mock interview. Only interviews in "scheduled" status can be cancelled.

Parameter	Type	Description
`interview_id`	string, required	The interview UUID to cancel.

`get_account_info`

Show who is currently signed in — display name, auth provider, and selected interests.

No parameters required.

Authentication

Skilark uses OAuth 2.0. On first use, your AI assistant will open a browser window where you sign in with GitHub, Google, or email magic link. After that, your session stays active automatically.

Scopes granted: read (job data), profile (career preferences), watch (job alerts).
Data stored: Email and display name from your OAuth provider. See our privacy policy.
Revoking access: Disconnect the connector in your AI assistant's settings.

Limitations

Market data tools are read-only. Watch, interview, and account tools can modify state.
Skills must match the canonical taxonomy for browse_jobs.
find_my_fit is slower than browse_jobs — vector search under the hood.
Up to 5 active watches per account. quick_chat interviews limited to 1 per day.