We build open-source tooling for people who run AI agents on their own infrastructure, we do longitudinal analysis of market data, and we run free services such as a rental-scam checker. Three lines of work, one standard: no accounts, no tracking, and no number that doesn't hold up.
Check a rental listing in Spain before you pay. Paste the listing and your chat with the “landlord” (or upload screenshots) and get a green, amber or red light on whether it looks like known scams, and why. Free, no sign-up.
Audits MCP configuration files for hardcoded keys, plaintext transports, curl | sh launch commands and over-broad permissions. It reads the configuration, not the servers. SARIF output.
pipx install mcp-config-auditPyPI is the only registry it is published to, always from a tagged commit in this repository. A package of this name on npm or any other registry is not ours.
Grammarly-style writing assistant for Chrome that uses your own API key, or a model on your own computer with no key: Chrome's built-in one, or one ProofKey runs inside the browser. No backend, no account, no telemetry. Tested with Gemini, xAI, OpenRouter, OpenCode Go and a local llama.cpp server.
Turns an epub into a chaptered audiobook. Gemini TTS does the narration; voice conversion against a single reference clip then anchors the timbre, so the voice stops changing every time the model starts a new generation. It ships with the measured research behind it: engine defects, blind listening rounds, and the fixes that improved the numbers while sounding worse.
All the code is at github.com/jiru-labs.
We track the residential supply of Madrid and its metropolitan area daily, listing by listing, from the day it is published until it disappears. That series yields figures an average-price report cannot give: how long each zone actually takes to sell, what share of sellers end up cutting the price, by how much, and how long they wait before doing it.
Time on market is estimated with survival analysis: listings still active are not discarded, they inform the estimate up to the last observed date. That is what avoids the optimistic bias of averaging only what has already sold. Aggregates are cross-checked against the INE's official transaction statistics for the province.
Figures as of 15 October 2026.
Delivered as a PDF report with the per-zone analysis, plus the working data in Excel. A one-off study over the accumulated history, or a quarterly update.
To request an extract for a zone, discuss a commission, or anything about the projects: info@jirulabs.com
Madrid, Spain