Gas
The first time Slime sees a request, an agent writes its endpoint on the spot and checks it. Unused ones evaporate.
Your app asks for data in plain words. Slime writes the endpoint, checks it, and returns results in the shape you want.
How it works
Describe the data you want in one GET to /slurp. The first wish builds the endpoint in a few seconds; every repeat comes from memory in milliseconds.
GET /slurp ?want=top N accounts by total paid invoices, highest first &limit=3 &shape=[{ name: string, totalPaidCents: integer }]
-- Slime will synthesize a safe, read-only
-- query here on the first request.
// response appears here
Matter states
Every endpoint starts as a first guess and hardens into permanent code, shaped only by what your product actually asks for.
The first time Slime sees a request, an agent writes its endpoint on the spot and checks it. Unused ones evaporate.
Seen it before? Slime finds the endpoint by its hash and serves it from memory, no AI involved. A second model reviews it before it's kept.
Endpoints that run hot crystallize: an agent optimizes them into ordinary code, committed to your repo.
Use cases
Coding agents write endpoints faster, but each one is still a ticket, a review and a deploy. With Slime, the request is the spec: the endpoint grows itself at runtime.
Vibecode the UI. Let Slime build the backend.
Point your generated UI at Slime. Every new widget gets its endpoint the moment it asks.
Unleash your agents. Keep your database untouchable.
Agents ask for the data they need and Slime builds the tool on the spot: read-only, reviewed, guarded by your database.
Don't rewrite legacy. Let Slime digest it.
Slime sits in front of old databases and services and grows a modern API on top, one request at a time.
One runtime. Every customer's backend.
Each customer's custom report becomes its own self-built endpoint, on one shared runtime.
Why Slime
Static APIs were built for software that changed slowly. Slime grows its API from what your product actually asks for.
Ship at the speed you design. New data features go live as fast as the UI that needs them.
End the backend-for-frontend grind. No more ticket queues for one more field in one more endpoint.
Move at AI speed, safely. Your database, not the AI, decides what Slime can touch.
Stop writing endpoints.Slime writes them.
Point your UI or agent at Slime. The backend builds itself.