← Blog

The SpaceX API Went Dark. We Put the Data Back Online.

api.spacexdata.com was archived in June and its origin now returns TLS errors — breaking thousands of tutorials, dashboards, and demo apps. We rebuilt its datasets from the original upstream sources, current again and synced 4x daily. Free REST routes + MCP.

For years, the community-run SpaceX API (api.spacexdata.com, the r/SpaceX-API project) was one of the most-used demo APIs on the internet. Coding bootcamps taught fetch() with it. GraphQL tutorials wrapped it. Portfolio dashboards, Postman collections, React courses — thousands of them point at those endpoints.

In June 2026 the project was archived, and the origin now returns a TLS 525 error. Every one of those tutorials and apps broke.

We noticed because our telemetry lit up: the old API was briefly the #1 error source across the whole Pipeworx catalog — hundreds of failed calls a week from AI agents still trying to reach it. The demand didn’t die with the hosting.

The interesting discovery

Reading the archived project’s source, the “SpaceX API” was never a primary source — it was an aggregator. Its sync jobs pulled from Launch Library 2 (launches, rockets, crew) and space-track.org (Starlink orbital elements), reshaped into its own schema. There is no official SpaceX API, and there never was.

The sources are alive. Only the hosting died. So we rebuilt it.

What’s live now

Free REST routes, no key, CORS-open, backed by a mirror we sync from the original upstream sources four times a day:

curl https://gateway.pipeworx.io/spacex/v4/launches/latest
curl https://gateway.pipeworx.io/spacex/v4/launches/next
curl https://gateway.pipeworx.io/spacex/v4/launches/past?limit=5
curl https://gateway.pipeworx.io/spacex/v4/rockets
curl https://gateway.pipeworx.io/spacex/v4/crew
curl https://gateway.pipeworx.io/spacex/v4/starlink?limit=10

GET /spacex/v4 returns the route index. Arrays accept ?limit=.

Two honest caveats:

  1. Route paths mirror the old v4 API; response schemas don’t. Our payloads are simplified rebuilds (e.g. rocket is the vehicle name, not a UUID you have to join). If your code parsed specific v4 fields, expect small changes. We chose readable-now over bug-compatible.
  2. In exchange, the data is current again. The original API’s data had gone stale long before the hosting died — its last consistent updates were years old. This mirror tracks LL2 and space-track.org, so launches/next is the actual next launch, and Starlink positions come from authoritative orbital data.

For AI agents

The same data is a first-class MCP pack — https://gateway.pipeworx.io/spacex/mcp — alongside 1,100+ other live sources on the gateway. Agents were the first users to feel the outage; now get_next_launch just works, with the mirror absorbing upstream rate limits so your agent never sees a 429 from LL2.

The pattern

This is a repeatable move we’re leaning into: watch what agents and apps keep asking for after an upstream dies, check whether the dead API was secretly an aggregator, and rehost the schema on infrastructure someone is paid to keep up. Dead APIs leave real demand behind — by one directory’s measure, roughly one in five MCP servers in the public registry no longer responds. The data usually outlives the hosting.

If there’s a dead API you miss, tell us — it might be one telemetry review away from coming back.