OSIRIS: sixteen live OSINT layers, two of them hand-maintained, and a scanner you wire up yourself
Open Source Global Intelligence Platform - Real-Time OSINT Dashboard - A Palantir Alternative - 2nZNHm3Lr9umG3DVrzYwHgktwkuKuJRXqqRqs3ewpump
At a glance
- What is it?
- OSIRIS is a MIT-licensed TypeScript OSINT dashboard that puts flight tracking, 17,000 public cameras, seismic events, fires, satellites, live news and sanctions search on one WebGL map, with a bring-your-own-key forecasting panel layered on top. The layers that look most authoritative are the ones that read least like live data.
- Who is it for?
- OSIRIS fits an analyst who wants one screen for flights, public cameras, seismic events and sanctions lookups, and who accepts that the maritime and conflict layers are hand-maintained snapshots sitting beside genuinely live feeds.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 21, 2026, and from our analysis. They are not legal advice.
Star history
- simplifaisoul/osiris11,832 ★ · +2,009
| Repository | 2026-09-21 | 2026-11-21 |
|---|---|---|
| simplifaisoul/osiris | 9.8k | 11.8k |
2,009 Stars in the last 61 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
Sixteen layers, and two of them are static snapshots
OSIRIS presents 16 toggleable data layers with live entity counts, and the table under the header is the honest way to read what feeds them. Aviation draws commercial, private, military and jet aircraft from the OpenSky Network, seismic comes from the USGS Earthquake API at M2.5 and above, fires from NASA FIRMS hotspots, space from NOAA SWPC and N2YO, news from 23 global broadcasters around the clock, crypto from blockstream.info, Blockscout and OpenSanctions, and sanctions is a person, organisation and vessel search against a mirror of the US OFAC SDN list. Two entries are hand-maintained rather than live feeds. Maritime is 39 global ports and 10 chokepoints labelled static naval intel, and conflict is 15 curated zones from a hand-kept list, shown alongside headlines from BBC, Al Jazeera and NYT world feeds. That distinction decides how you should read the globe, because a marker in either of those two layers is as current as the last edit that touched those files while a flight on the aviation layer actually moves. Treating all 16 as the same class of evidence is the fastest way to draw a wrong conclusion from a screen that looks uniform.
Seventeen thousand cameras, every one of them a JPEG snapshot
The CCTV layer is the largest by count, at 17,000 plus cameras, drawn from transport and highway authorities including TfL, WSDOT, Caltrans, ODOT, MDOT, the Hong Kong Transport Department, Taiwan THB, NZTA and Rijkswaterstaat, plus public webcams. How current any of them are depends on what the source will give you, and the Texas CCTV section states the limit most explicitly. Public TxDOT ITS snapshots are integrated with the existing camera markers, preview grid and viewer, and the note is blunt: availability varies by district and camera, and these are refreshing JPEG snapshots, not video streams. Texas alone is reachable with /api/cctv?region=texas, while global and Texas location queries include the same source. District inventories are cached independently and stale data is deliberately retained during outages, so a district going dark leaves its last frame sitting in the map rather than clearing it. The consequence is that, for the Texas snapshots, a frozen image and a live one can look alike.
Drawing a region only exports cameras, and the named endpoints are few
The one export path the documentation walks through is geometric. You draw a region on the map and OSIRIS finds every camera inside it, ready to export as GeoJSON, which is what the Save an Area screenshot demonstrates. Beyond that, the programmatic surfaces the README names are a REST API under /api/oi with live Server-Sent Events, an MCP server at /api/mcp so agents such as Hermes, Claude and Cursor can forecast, steer runs and read OSIRIS intelligence, and the camera query above. The architecture diagram lists Next.js API routes such as /api/flights and /api/earthquakes, and the drawn region is scoped to cameras rather than acting as a general extraction feature. Read the documentation anchor linked from the feature list before you assume a bulk export path exists, because the interactive map and the API are not the same surface.
The map is a WebGL canvas, so nothing in it is inspectable
Rendering is the architectural decision with the widest consequences. All map data is drawn through WebGL rather than DOM, inside MapLibre GL, and the stated goal is 60fps with thousands of concurrent entities on screen. Data is fetched progressively when a layer is activated, and it is viewport-aware, so only what is visible is loaded. The cost is that the map is a canvas: the flights, cameras and zone polygons you are looking at are not document nodes, so find-in-page, text selection and DOM-scraping extensions see nothing, and any automation written against marker elements has nothing to query. The quantity worth watching instead is the per-layer entity count, since that is what the renderer is actually holding, and the toggle list is how you keep that number honest on a slow machine. Progressive loading also means a layer can look empty while its first request is still in flight.
The RECON panels return 503 until a scanner backend is configured
The RECON toolkit, port scan, DNS and WHOIS, vulnerability scanner, is the one part of OSIRIS that does not work out of the box. The environment template is explicit that the application code currently reads only SCANNER_URL and SCANNER_KEY, that this pair powers the quick, ssl, headers, rdns, subdomains, tech, whois, geoloc and vuln scans, and that leaving both empty simply disables the feature with the UI returning 503. Behind it sits a separate scanner backend you run yourself, not a hosted service, and the two halves must agree on a shared secret: SCANNER_KEY has to equal the backend's OSIRIS_KEY, generated with:
openssl rand -hex 32The compose file does start a service built from the intel directory on port 4000, yet the template ships SCANNER_URL empty, so the two are not connected until you fill it in. Everything else works keyless, which the template also states: aviation, maritime, satellites, fires, earthquakes, weather, news and CVEs all use public feeds including adsb.lol, celestrak.org, the NASA FIRMS open CSV and USGS.
The environment file gives two contradictory answers about the Cloudflare token
The header of the environment template says OSIRIS works fully without any third-party API keys, then says the application code only actually reads SCANNER_URL and SCANNER_KEY, with every other key optional and reserved for higher rate limits or future data sources. There is a mismatch: the capabilities table credits aviation to the OpenSky Network while the template names adsb.lol among the keyless feeds.
A 5.0 release tag on a manifest that still reads 0.1.0 private
The release history and the package manifest do not line up. The newest release is v5.0.0, tagged OSIRIS V5.0 on 2026-09-28, and the last push to master is 2026-11-02, so the tree is moving faster than the tags. Meanwhile package.json names the project osiris at version 0.1.0 with private set to true, which means there is no registry package to install and no version field that tells you which build you cloned. You get the source tree, the container, or the live demo. The container path is the one the repository supports: the compose file builds locally from the Dockerfile so you always run the cloned code, and its own comment gives the substitution for the prebuilt image:
image: ghcr.io/simplifaisoul/osiris:latestThe host port comes from OSIRIS_PORT in .env and defaults to 3000, the cache tier is nginx on 8080 with its config mounted read only, and the tree also carries DOCKER.md and deploy.sh. Verifying a running instance before trusting it means checking which of the two public addresses you are on, since the badge in the README points at osirislive.app while the demo and the docs both sit on osirisai.live.
Editorial conclusion
OSIRIS fits an analyst who wants one screen for flights, public cameras, seismic events and sanctions lookups, and who accepts that the maritime and conflict layers are hand-maintained snapshots sitting beside genuinely live feeds. It fits badly as a system of record, because nothing in the manifest or the release tags identifies which build you cloned, and the Texas camera feed is refreshing JPEG snapshots rather than video, with stale data retained during outages, so a frozen frame can pass for a current one. Before deploying, confirm two things: which of the two public domains you trust, and whether you actually need the RECON backend wired to port 4000.
Frequently asked questions
how to install osiris
Clone the repository and use the container path: the compose file builds from the Dockerfile so you run the cloned code, and the host port is set with OSIRIS_PORT in .env, defaulting to 3000. No third-party API key is required, since the public feeds are keyless.
how to access osiris
A live demo runs at https://osirisai.live with documentation at https://osirisai.live/docs#oi. Self-hosted it serves on port 3000 by default, exposes a REST API under /api/oi with live Server-Sent Events, and an MCP server at /api/mcp, and returns Texas cameras from /api/cctv?region=texas.
how to use osiris
Toggle the 16 data layers, which load progressively and only fetch what is in the viewport. Press O for OI Assist, ask in words typed or spoken, and it flies the map to the answer, switches layers on and marks what it found in cyan.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/simplifaisoul-osiris)Star history chart for your README
[](https://hysenlabs.com/projects/simplifaisoul-osiris)Was this useful?