Where this stands
A working thing, not a service
Adeia is open source and self-hosted. There is no signup, no dashboard and no hosted API — the server runs on your machine, the database is a file next to it, and the API key comes from a script you run yourself. That is the whole distribution model today, and it is stated here rather than discovered after you try to register.
01 — State
Real, and deliberately unfinished
Built and working
- A policy engine that decides allow, hold or deny with the figure that produced it
- Email approvals with single-use, expiring, hashed tokens
- An append-only audit log with redaction at write time
- A TypeScript SDK — three methods, no hidden retries
- 293 tests, and two full runs against real mail and a real phone
Deliberately not built
- No payment processor. The adapter seam is empty on purpose — payments are recorded, nothing settles
- One approver per deployment, from an environment variable
- SQLite, so one writer and one machine
- No rate limiting, and no API key rotation
- @adeia/sdk is not published to npm; it ships in the repository as a workspace
Built for the Lumos Fellows program. Every layer above the adapter is real; the seam below it is the part left open on purpose, because a convincing fake payment processor is worse than an obvious gap.
02 — Involved
The repository is the channel
There is no inbox behind this project yet, so rather than print an address that bounces: issues and pull requests are where anything actually gets discussed.
Run it and say what broke
The most useful thing anyone can do right now. Clone it, seed a project, point an agent at it, and open an issue when the instructions are wrong or something fails in a way the docs did not predict. It has been run by very few people.
Attach an adapter
The registry takes any action type, not just payments — sending mail, deleting records, calling a paid API. A second adapter would prove the seam is real rather than theoretical.
Break the approval flow
Tokens are hashed, single use and expiring, and nobody outside this project has tried to defeat that. A serious attempt would be worth more than a feature.