AT Crab#

Rust AT Protocol libraries.

Status: This is a personal project hacked together to scratch an itch. It will remain v0.x until the API is stable, test coverage is high, and I'm confident others can depend on it. At that point I plan to publish to crates.io. AI agents contributed to the implementation, but I've steered the API design toward idiomatic Rust and a human-ergonomic developer experience.

Do not use this in production. See https://atproto.com/sdks for alternatives.

Features#

• Handle & DID resolution - resolve any handle to a DID via DNS (_atproto) or .well-known, then resolve the PDS endpoint from PLC or did:web documents
• Read records - fetch records from any collection with built-in pagination and automatic cursor handling
• Write records - create, update, and delete records in your own PDS repository
  • Auth - login with app passwords, auto-loaded from .env or environment variables (ATP_PASSWORD)
• Blob upload - upload images and other binary data to your PDS
• Built-in lexicon types - ready-to-use types for standard.site lexicons

Adding as a dependency#

[dependencies]
atcrab = { git = "https://github.com/metruzanca/atcrab" }

Usage#

Fetch with built-in types#

Built-in lexicon types implement Collection so the NSID is inferred:

use atcrab::lexicons::Document;
use atcrab::Repo;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let repo = Repo::new("metru.dev").await?;
    let docs = repo.fetch_collection::<Document>().await?;
    for record in &docs.records {
        println!("{} ({})", record.value.title, record.uri);
    }
    Ok(())
}

Included types:

Shared ATProto types: Blob, BlobLink, StrongRef, SelfLabel

Fetch with custom types#

Implement Collection on your own types for the same inferred API:

use atcrab::{Collection, Repo};
use serde::Deserialize;

#[derive(Deserialize)]
struct Post { text: String }

impl Collection for Post {
    const NSID: &'static str = "app.bsky.feed.post";
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let repo = Repo::new("metru.dev").await?;
    let posts = repo.fetch_collection::<Post>().await?;
    println!("{} posts", posts.records.len());
    Ok(())
}

Low-level API#

If you don't want to implement Collection, pass the NSID directly:

// first page
let posts = repo.fetch::<Post>("app.bsky.feed.post").await?;
// next page from cursor
let posts = repo.fetch_cursor::<Post>("app.bsky.feed.post", cursor).await?;
// all pages
let all   = repo.fetch_all::<Post>("app.bsky.feed.post").await?;

Examples#

cargo run --example basic           # fetch documents with built-in types
cargo run --example custom_schema   # fetch posts with a user-defined type
cargo run --example pagination      # cursor-based pagination
cargo run --example render          # pretty-print documents