Authentication¶
The API is public and keyless. Every endpoint works with no credentials. Just send the request. This is the same access the web app at japanfold.com uses, with the same limits.
# No auth needed:
curl -s https://api.japanfold.com/v1/models
Optional API key: raised limits¶
Send a key of the form jf_live_… to lift the free-tier caps (larger
structures, more chains, more jobs at once), as Authorization: Bearer or
X-API-Key:
curl -s https://api.japanfold.com/v1/predictions \
-H 'Authorization: Bearer jf_live_your_key_here' \
-H 'Content-Type: application/json' \
-d '{"model":"boltz2","sequence":"MKTAYIAK..."}'
A key that's present but invalid is rejected with 401. Omit the header
entirely to fall back to the public tier. Keys are issued by the JapanFold
team; there's no self-serve signup yet.
Ownership and quotas¶
- Keyless requests are grouped and rate-limited per client IP (and per session for the web app). This is a free demo on shared compute, so there are quotas on how much you can run at once.
- Keyed requests are owned by your key. You can only list, poll, cancel or
delete jobs that belong to you (your IP/session for keyless use, your key
otherwise). Accessing someone else's job returns
404. - The concrete numbers (active-job quotas, submit rates, structure/design size
caps) are in Models & limits and, live, at
GET /v1/models.
Over a cap you get 400; at capacity you get 429 with a Retry-After header
(see Errors).