repome

The public API is the same typed contract used by the CLI and SDK. The RPC transport lives at:

https://api.repome.sh/api/rpc

An OpenAPI view is also published for HTTP clients:

https://api.repome.sh/api/v1/openapi.json
https://api.repome.sh/api/v1/docs

Authentication

Signed-in requests use an API key:

Authorization: Bearer rpme_...

Create one with the CLI:

repome keys create ci --scope repos:read --scope repos:write

Anonymous repo requests use an anonymous identifier:

X-Repome-Anon-Id: <id>

The high-level repome.ephemeral() and repome anon commands create and reuse that identifier for you.

Procedure groups

Group	Purpose
`health.ping`	API health probe
`me.get`	Current user, active organization, and membership context
`orgs.*`	Create, list, get, and delete organizations
`members.*`	List, invite, remove, accept, and update organization members
`repos.*`	Create, import, ensure, get, list, update, and delete repos
`git.*`	Read refs, commits, trees, blobs, compare/diff, and commit workspace changes
`tokens.*`	Mint, list, revoke, and validate repo-scoped git tokens
`keys.*`	Create, list, and delete API keys
`anon.*`	Create and manage anonymous ephemeral repos and their git reads/writes

Repos are addressed by bare name everywhere — repos.*, git.*, and tokens.* inputs carry no org field; the server resolves names against the caller's active organization. orgs.create takes only an optional name: the slug is server-minted (an opaque org-… handle). orgs.delete refuses to delete your last owned org (ORG_LAST_OWNED) or an org that still owns repos (ORG_HAS_REPOS).

Raw file/blob route

For fetching file bytes directly — downloads, browser viewers, curl — use the raw byte route instead of the base64-wrapped git.blob.get procedure:

GET  /:name/raw/:ref/<path>
HEAD /:name/raw/:ref/<path>

:name is the bare repo name, resolved against the session's active organization. :ref is a single segment — a branch name, HEAD, or a full commit sha (a short branch name is expanded to refs/heads/<name>). The response is the raw bytes (no JSON/base64 envelope) with Content-Type: application/octet-stream, Content-Length, ETag (the blob oid), X-Repome-Blob-Oid, and X-Repome-Blob-Size. HEAD returns the same headers with no body.

curl -H "Authorization: Bearer rpme_..." \
  https://api.repome.sh/site/raw/main/public/logo.png -o logo.png

Responses are bounded at 10 MiB. A larger blob returns 413 Payload Too Large with X-Repome-Blob-Limit (and X-Repome-Blob-Oid / X-Repome-Blob-Size).

Raw read vs. git clone/fetch

Raw route / git.blob.get — single files you can name by path, up to the limits above (git.blob.get returns base64 JSON capped at 1 MiB; the raw route returns bytes up to 10 MiB). Best for previews, downloads, and tooling that already knows the path.
git clone / git fetch — whole repositories, history, or any object past the raw limit. Large files always belong on the git-native transport, which streams from Artifacts without materializing the object in one piece.

Low-level SDK client

For TypeScript callers, @repome/sdk exposes the raw typed RPC client:

import { createClient } from '@repome/sdk'

const client = createClient({
  baseUrl: 'https://api.repome.sh',
  apiKey: process.env.REPOME_API_KEY!,
})

const repos = await client.repos.list({})

Use the OpenAPI document when generating non-TypeScript clients.

CLI reference

Manage the same API from a shell.

Git SDK

Use the high-level JavaScript git API.

API reference