Documentation

Docs

Everything here assumes you have a token. If you do not, take one from the home page — it takes a click and no sign-up.

Start here

• Quickstart — token to first retrieved bundle, in about ten minutes.
• API reference — every endpoint, error code, and header.
• Beta limits — what the quotas are and what happens when you reach them.

How it fits together

rebornpad is a place to put the working memory of a coding agent so it survives between runs. You chunk your source, embed it with whatever model you already use, and stream both into a collection. Later the agent searches that collection and gets back a single packed bundle — matching chunks, their metadata, and the state it saved last time.

We deliberately do not embed anything for you. Bring your own model, keep your own dimensions, and swap models without asking us. The trade-off is that re-embedding means re-ingesting.

What it is not

• Not a code host. We store chunks you send, not repositories, and there is no git integration.
• Not an inference service. No models run here.
• Not a general-purpose vector database. The schema is fixed around agent context; if you need filters, hybrid search and sharding, use a real vector store.

Getting help

There is no support desk. Security reports go to the address in security.txt. For anything else, the changelog is usually where the answer to "did something change?" lives.