SongSage

SongSage is a Model Context Protocol (MCP) server that connects BirdNET-Analyzer-Sierra with Claude Desktop, enabling natural language interaction with bioacoustic data for wildlife monitoring and conservation research.

Why SongSage?

Bioacoustic monitoring is a powerful tool for studying biodiversity, but BirdNET outputs are static CSV files requiring custom scripts to analyze. SongSage transforms these detections into an interactive, conversational analysis system.

Ecologists, conservation practitioners, and citizen scientists can now query, summarize, and visualize acoustic data using natural language—no coding required.

Inspired by multimodal wildlife monitoring research, including the SmartWilds framework at The Wilds Conservation Center.

Architecture

flowchart TB
    subgraph Users
        ECO[Ecologists]
        CIT[Citizen Scientists]
    end

    subgraph Interface
        CD[Claude Desktop]
        MCP[SongSage MCP Server]
    end

    subgraph Pipeline
        AF[Audio Files]
        BN[BirdNET-Analyzer-Sierra]
        DR[Detection Results]
    end

    subgraph Analytics
        QRY[Query Engine]
        VIZ[Visualization]
    end

    ECO --> CD
    CIT --> CD
    CD --> MCP
    AF --> BN
    BN --> DR
    DR --> MCP
    MCP --> QRY
    MCP --> VIZ
    QRY --> Users
    VIZ --> Users

Key Capabilities

Installation

See docs/installation.md for the complete setup guide, including platform-specific configuration for Mac/Linux and Windows.

Quick summary:

1. Clone the repo and run setup.sh (Mac/Linux) or set up the venv manually (Windows)

2. Add SongSage to your Claude Desktop config with full absolute paths

3. Optionally create a .env file if BirdNET isn't auto-detected

Usage Examples

Daily Monitoring

"Summarize bird activity from today's recordings."

Rare Species

"Find species detected fewer than 3 times with confidence above 0.7."

Peak Activity

"When are birds most active during the day?"

Species Deep Dive

"Show me everything about Northern Cardinal detections."

Temporal Comparison

"Compare bird activity between June and July."

Visualization

"Generate a heatmap of activity by hour for the top 10 species."

Tools

Analysis

Queries

Visualization

Utilities

Guided Workflows (Prompts)

Pre-built multi-step analyses:

Project Structure

SongSage/
├── assets/               # Logo 
├── docs/                 # Documentation
│   ├── documentation.md  # Full technical reference
│   └── installation.md   # Installation and configuration guide
├── heatmaps/             # Generated visualizations
├── test_data/            # Sample CSV files for setup verification (see installation guide)
├── mcp_server.py         # MCP server implementation
├── requirements.txt      # Python dependencies
├── __init__.py           # Python package init
├── .env.example          # Configuration template
└── setup.sh              # Linux/macOS installer

Research Context

SongSage builds on multimodal wildlife monitoring approaches. The SmartWilds project demonstrates how bioacoustic sensors complement camera traps and drone imagery for ecosystem monitoring—bioacoustics provide continuous temporal coverage and detect species that visual methods miss.

This tool lowers the barrier for researchers and citizen scientists to explore acoustic biodiversity data through conversation rather than code.

Acknowledgments

• BirdNET Team — Cornell Lab of Ornithology & Chemnitz University

• Anthropic — Claude and Model Context Protocol

• The Wilds Conservation Center

• Imageomics Institute & ABC Global Center

• SmartWilds Team, The Ohio State University and The Wilds

This work was supported by both the Imageomics Institute and the AI and Biodiversity Change (ABC) Global Center. The Imageomics Institute is funded by the US National Science Foundation's Harnessing the Data Revolution (HDR) program under Award #2118240 (Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning). The ABC Global Center is funded by the US National Science Foundation under Award No. 2330423 and Natural Sciences and Engineering Research Council of Canada under Award No. 585136. This project draws on research supported by the Social Sciences and Humanities Research Council. Some additional support was provided by the NSF AI Institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE), funded under NSF Award OAC-2112606. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation, Natural Sciences and Engineering Research Council of Canada, or Social Sciences and Humanities Research Council.