An AI agent skill that pinpoints photo locations down to precise coordinates using spatial geometry and terrain data.
Pinpoint ungrounded photos using elevation skylines, OpenStreetMap corridors, and satellite imagery.
geo-sleuth is an agent skill that automates visual geolocation for unindexed or metadata-stripped photographs. It equips CLI and IDE AI agents like Claude Code or Cursor with Python scripts to calculate terrain skylines, shadow angles, and OpenStreetMap geometry. Photographers and researchers can use it to pinpoint exact shooting locations without relying on EXIF data.
The skill executes a multi-step analytical workflow combining metadata extraction, terrain elevation matching, satellite tile scanning, and skyline fitting. Rather than relying on LLM visual guessing, 20 dedicated Python scripts filter and rank candidate sites before passing top matches to the model. Output reports include camera coordinates, heading direction, estimated error radius, and satellite evidence images.
This tool suits technical photographers, forensic image analysts, and OSINT researchers who already work with AI CLI agents and need precise spatial context for landscapes or field photos. It is not suitable for casual users looking for a standalone graphical application or instant drag-and-drop web app.
Renders candidate ridge lines from elevation data to fit mountain profiles found in photo backgrounds.
Queries structural geometry like railway bridges, pier spacings, and roads from OpenStreetMap to narrow candidate regions.
Runs standard SKILL.md and Python scripts directly inside Claude Code, Cursor, Codex, Gemini CLI, or GitHub Copilot.
Tracks hypotheses using likelihood ratios and forces every location claim to reference generated output files.
geo-sleuth is free and open-source under the MIT License. The only costs involved are your local machine's processing power and API usage from your underlying host AI agent.
880 open remote AI-training roles, updated November 16, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
No, geo-sleuth is specifically designed for photos where EXIF metadata has been stripped, relying on physical visual features and spatial terrain matching.
It supports any agent that reads SKILL.md and executes shell commands, including Claude Code, Cursor, Gemini CLI, OpenCode, Codex, and GitHub Copilot.
Simple metadata checks finish in seconds, but complex regional skyline and OpenStreetMap scans can take 30 to 70 minutes of local computation.
See Zekai first in Google