Skip to main content
Glama
Apartment Hunter MCP Server

Apartment Hunter MCP Server

Apartment Hunter

WARNING

Legal Disclaimer: This is a private automation tool for personal use only. By using this software, you agree to the Legal Disclaimer and Fair Use Policy.

Apartment Hunter is a powerful aggregator and search platform for rental apartments. It automates finding, filtering, and scoring housing options without manually checking websites every hour. Currently, the system only supports data from krisha.kz, but the architecture allows for easy extension. It supports precise polygon-based geographic search, semantic text search (RAG), and LLM-powered quality scoring.

Quick Start

# 1. Start the MCP server for Codex Desktop / Claude
uv run apartment-hunter-mcp

# 2. Or run the data ingestion pipeline manually
uv run apartment-hunter-ingest

AI Skills

The project includes specialized AI skills in .agents/skills to enhance your AI assistant:

  • apartment-search: Domain workflow for apartment search, ingestion, monitoring, and translating listing jargon.

  • city-district-context-kz: Provides context on Kazakhstan cities and districts (transport, pricing, livability).

  • html-report-generator: Generates a visual HTML report of apartments with photos and links, and launches it on localhost.

How to use skills: You don't need to manually install these skills. Simply open this project folder using your favorite AI CLI (like Antigravity CLI / Codex CLI) and the agent will automatically discover and load the skills from the local workspace.

Related MCP server: @striderlabs/mcp-airbnb

Use Cases

  • Agents / MCP: Connect it to Codex Desktop to chat with your apartment database ("Show me 2-bedroom apartments near Megapark under 500k, sort by newest").

  • Precise Geography: Stop relying on inaccurate text search. Draw a polygon on a map, pass the coordinates, and only get apartments strictly within that exact boundary.

  • AI Scoring: Automatically rate apartments out of 10 based on renovation quality, furniture condition, and specific user preferences using Vision LLMs.

  • Market Monitoring: Run ingestion periodically (e.g. via cron) to track price drops and immediately catch new listings before they are rented out.

Architecture Philosophy

This is a Library-First project. The core logic is built as a reusable, source-agnostic Python library. All site-specific scraping and adapters are isolated plugins, ensuring the core remains stable and extensible.

License

MIT

Available Tools

15 tools
analyze_apartmentA

Run LLM analysis on a specific apartment.

Scores the apartment 0-10 based on price/quality ratio, condition, location, and other factors. Returns score, pros, cons, and summary. Forces re-analysis even if already analyzed.

Inputs: source_id (required). Returns: result

compare_apartmentsA

Compare 2-5 apartments side by side.

Provide a list of source_ids to compare their key characteristics.

Inputs: source_ids (required). Returns: result

create_search_profileB

Create a search profile for ongoing apartment monitoring.

The profile defines filters and preferences. The ingestion pipeline will use these profiles to fetch new apartments and send notifications.

Inputs: name (required), city, rooms, sources, +12 more. Returns: result

delete_search_profileC

Delete a search profile by ID.

Inputs: profile_id (required). Returns: result

download_apartment_photosA

Download apartment photos locally so the AI agent can inspect them.

Returns the absolute paths of the downloaded images. You (the AI) can then use your view_file tool on these paths to visually analyze the apartment.

Inputs: source_id (required), limit. Returns: result

get_apartment_detailsA

Get full details for a specific apartment by its source_id (e.g. 'krisha:1013405508').

Returns all available fields including LLM analysis, price history, and photos.

Inputs: source_id (required). Returns: result

get_new_apartmentsA

Get apartments discovered in the last N hours (default: 24).

Shows the most recent apartments sorted by score.

Inputs: since_hours. Returns: result

get_price_historyC

Get price change history for an apartment.

Inputs: source_id (required). Returns: result

get_statsA

Get statistics about the apartment database.

Shows total apartments, new apartments, analyzed count, average prices, scores, etc.

Inputs: no named parameters. Returns: result

get_top_apartmentsA

Get top-rated apartments sorted by LLM score.

Returns the highest-scored apartments, optionally filtered by city, rooms, or price. Only returns apartments that have been analyzed.

Inputs: city, limit, rooms, price_max. Returns: result

list_search_profilesA

List all active search profiles.

Inputs: no named parameters. Returns: result

run_ingestionA

Run the data ingestion pipeline.

Fetches new apartments from all sources for the specified profile (or all active profiles if none specified), analyzes them, and sends notifications. This may take several minutes depending on the number of pages to scrape.

Inputs: profile_id. Returns: result

search_apartmentsA

Search apartments by structured filters.

Returns a list of apartments matching the given criteria. Use this for precise filtering by price, rooms, area, district, etc. For free-text queries like 'cozy apartment near park', use semantic_search instead.

Inputs: city, rooms, area_max, area_min, +5 more. Returns: result

search_by_profileA

Show apartments matching a search profile, ranked by semantic similarity.

Uses the profile's hard filters (price, rooms, polygon) AND the nl_description for semantic re-ranking via ChromaDB vector search. Pass hours=0 to return all stored apartments for this profile (no time filter).

Inputs: profile_id (required), hours, limit. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.5/5.0

    Scored across 15 tools

    Disambiguation4/5

    Most tools have distinct purposes, but `semantic_search` and `search_apartments` could be confused without reading descriptions. Also `analyze_apartment` and `get_apartment_details` overlap slightly.

    Naming Consistency4/5

    The majority follow a clear verb_noun pattern in snake_case. `semantic_search` is the main deviation, being adjective_noun, but it still reads naturally.

    Tool Count5/5

    With 15 tools, the server covers all major apartment hunting workflows (discovery, analysis, monitoring, comparison) without being overwhelming.

    Completeness4/5

    Core operations like search, details, analysis, and profile management are covered. Missing an update search profile tool is a minor gap, but the set handles the primary use cases.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      MCP server for Stessa rental-property accounting platform, enabling AI agents to query properties, portfolios, banking, documents, and tenancies via natural language.
      MIT
    • A
      license
      A
      quality
      D
      maintenance
      MCP server exposing portfolio AI tools including semantic search, evaluation framework, and prompt management, enabling natural language interaction with these services via Claude Desktop.
      5
      ISC
    • F
      license
      Not graded
      quality
      D
      maintenance
      A production-grade MCP server enabling Claude to perform comprehensive NJ real estate workflows including property search, valuation, neighborhood intelligence, investment analysis, and agent tools via 20 tools and 15+ data sources.
      -