diff --git a/upub/core/src/traits/fetch.rs b/upub/core/src/traits/fetch.rs
index c4a4b40..9081bc6 100644
--- a/upub/core/src/traits/fetch.rs
+++ b/upub/core/src/traits/fetch.rs
@@ -1,6 +1,6 @@
 use std::collections::BTreeMap;
 
-use apb::{Activity, Actor, ActorMut, Base, Collection, Object};
+use apb::{Activity, Actor, ActorMut, Base, Collection, CollectionPage, Object};
 use reqwest::{header::{ACCEPT, CONTENT_TYPE, USER_AGENT}, Method, Response};
 use sea_orm::{ActiveValue::Set, ColumnTrait, ConnectionTrait, DbErr, EntityTrait, IntoActiveModel, NotSet, QueryFilter};
 
@@ -401,9 +401,40 @@ impl Fetcher for crate::Context {
 		Ok(activity_model)
 	}
 
-	async fn fetch_thread(&self, _id: &str, _tx: &impl ConnectionTrait) -> Result<(), RequestError> {
-		// crawl_replies(self, id, 0).await
-		todo!()
+	async fn fetch_thread(&self, id: &str, tx: &impl ConnectionTrait) -> Result<(), RequestError> {
+		tracing::info!("crawling replies of '{id}'");
+		let object = self.pull(id).await?.object()?;
+		let replies = object.replies().resolve(self).await?;
+
+		let mut page;
+		let mut next = replies.first();
+
+		loop {
+			page = next.resolve(self).await?;
+
+			// fix for mastodon: at some point it introduces ?only_other_accounts=true and then returns a
+			// collection, not a page anymore ???
+			if matches!(page.object_type()?, apb::ObjectType::Collection(..)) {
+				page = page.first().extract().ok_or(RequestError::Tombstone)?;
+			}
+
+			for obj in page.items() {
+				if let Err(e) = self.fetch_object(obj.id()?, tx).await {
+					tracing::warn!("error fetching reply: {e}");
+				}
+			}
+
+			for obj in page.ordered_items() {
+				if let Err(e) = self.fetch_object(obj.id()?, tx).await {
+					tracing::warn!("error fetching reply: {e}");
+				}
+			}
+
+			next = page.next();
+			if next.is_empty() { break };
+		}
+
+		Ok(())
 	}
 
 	async fn fetch_object(&self, id: &str, tx: &impl ConnectionTrait) -> Result<crate::model::object::Model, RequestError> {
@@ -486,55 +517,3 @@ impl Dereferenceable<serde_json::Value> for apb::Node<serde_json::Value> {
 		}
 	}
 }
-// #[async_recursion::async_recursion]
-// async fn crawl_replies(ctx: &crate::Context, id: &str, depth: usize) -> Result<(), PullError> {
-// 	tracing::info!("crawling replies of '{id}'");
-// 	let object = crate::Context::request(
-// 		Method::GET, id, None, &format!("https://{}", ctx.domain()), &ctx.app().private_key, ctx.domain(),
-// 	).await?.json::<serde_json::Value>().await?;
-// 
-// 	let object_model = crate::model::object::Model::new(&object)?;
-// 	match crate::model::object::Entity::insert(object_model.into_active_model())
-// 		.exec(ctx.db()).await
-// 	{
-// 		Ok(_) => {},
-// 		Err(sea_orm::DbErr::RecordNotInserted) => {},
-// 		Err(sea_orm::DbErr::Exec(_)) => {}, // ughhh bad fix for sqlite
-// 		Err(e) => return Err(e.into()),
-// 	}
-// 
-// 	if depth > 16 {
-// 		tracing::warn!("stopping thread crawling: too deep!");
-// 		return Ok(());
-// 	}
-// 
-// 	let mut page_url = match object.replies().get() {
-// 		Some(serde_json::Value::String(x)) => {
-// 			let replies = crate::Context::request(
-// 				Method::GET, x, None, &format!("https://{}", ctx.domain()), &ctx.app().private_key, ctx.domain(),
-// 			).await?.json::<serde_json::Value>().await?;
-// 			replies.first().id()
-// 		},
-// 		Some(serde_json::Value::Object(x)) => {
-// 			let obj = serde_json::Value::Object(x.clone()); // lol putting it back, TODO!
-// 			obj.first().id()
-// 		},
-// 		_ => return Ok(()),
-// 	};
-// 
-// 	while let Some(ref url) = page_url {
-// 		let replies = crate::Context::request(
-// 			Method::GET, url, None, &format!("https://{}", ctx.domain()), &ctx.app().private_key, ctx.domain(),
-// 		).await?.json::<serde_json::Value>().await?;
-// 
-// 		for reply in replies.items() {
-// 			// TODO right now it crawls one by one, could be made in parallel but would be quite more
-// 			// abusive, so i'll keep it like this while i try it out
-// 			crawl_replies(ctx, reply.href(), depth + 1).await?;
-// 		}
-// 
-// 		page_url = replies.next().id();
-// 	}
-// 
-// 	Ok(())
-// }
diff --git a/upub/routes/src/activitypub/object/replies.rs b/upub/routes/src/activitypub/object/replies.rs
index 7a415ed..425df01 100644
--- a/upub/routes/src/activitypub/object/replies.rs
+++ b/upub/routes/src/activitypub/object/replies.rs
@@ -1,7 +1,7 @@
 use apb::{BaseMut, CollectionMut, LD};
 use axum::extract::{Path, Query, State};
 use sea_orm::{ColumnTrait, Condition, QueryFilter, QuerySelect, SelectColumns};
-use upub::{model, Context};
+use upub::{model, traits::Fetcher, Context};
 
 use crate::{activitypub::{Pagination, TryFetch}, builders::JsonLD, AuthIdentity};
 
@@ -9,11 +9,17 @@ pub async fn get(
 	State(ctx): State<Context>,
 	Path(id): Path<String>,
 	AuthIdentity(auth): AuthIdentity,
-	Query(_q): Query<TryFetch>,
+	Query(q): Query<TryFetch>,
 ) -> crate::ApiResult<JsonLD<serde_json::Value>> {
-	// if auth.is_local() && q.fetch {
-	// 	ctx.fetch_thread(&oid).await?;
-	// }
+	let oid = ctx.oid(&id);
+	if auth.is_local() && q.fetch {
+		// TODO a task should do this, not the web handler!
+		//      so we dont keep clients waiting and we limit
+		//      concurrent possible crawlers
+		//      however the results given immediately would
+		//      become inaccurate!!
+		ctx.fetch_thread(&oid, ctx.db()).await?;
+	}
 
 	let replies_ids = upub::Query::objects(auth.my_id())
 		.filter(auth.filter())