diff --git a/parser/src/earley/parser.rs b/parser/src/earley/parser.rs
index 58e7859c..90c36c8b 100644
--- a/parser/src/earley/parser.rs
+++ b/parser/src/earley/parser.rs
@@ -670,6 +670,33 @@ impl ParserState {
         }
     }
 
+    pub fn bytes_allowed(&mut self, shared: &mut SharedState, tok_bytes: &[u8]) -> bool {
+        self.assert_definitive();
+        let applied_idx = self.byte_to_token_idx.len();
+        let tok_bytes = if applied_idx < self.bytes.len() {
+            let prefix_len = std::cmp::min(tok_bytes.len(), self.bytes.len() - applied_idx);
+            if self.bytes[applied_idx..applied_idx + prefix_len] != tok_bytes[..prefix_len] {
+                return false;
+            }
+            &tok_bytes[prefix_len..]
+        } else {
+            tok_bytes
+        };
+        if tok_bytes.is_empty() {
+            return true;
+        }
+
+        self.run_speculative(|s| {
+            let mut r = ParserRecognizer { shared, state: s };
+            for &b in tok_bytes {
+                if !r.try_push_byte(b) {
+                    return false;
+                }
+            }
+            true
+        })
+    }
+
     // apply_tokens() "pushes" the bytes in 'tokens' into the lexer and parser.  It is a top-level
     // method in this file.  It is well below llguidance's top-level methods, but in the llguidance
     // LLInterpreter interface, it is called indirectly via the commit_token() method.
@@ -2001,6 +2028,11 @@ impl Parser {
         r
     }
 
+    pub fn bytes_allowed(&mut self, tok_bytes: &[u8]) -> bool {
+        let mut shared = self.shared.lock().unwrap();
+        self.state.bytes_allowed(&mut shared, tok_bytes)
+    }
+
     pub fn filter_max_tokens(&mut self) {
         self.state.filter_max_tokens();
     }