diff --git a/.config/nextest.toml b/.config/nextest.toml
index 4beb89f..f0f9abf 100644
--- a/.config/nextest.toml
+++ b/.config/nextest.toml
@@ -1,4 +1,7 @@
 # Default profile — used by `just test` and ad-hoc `cargo nextest run`.
+# `just test` applies `-E 'not binary(real_world_roundtrip)'` to filter
+# out network-dependent tests. Run them explicitly via
+# `just test-real-world` / `just test-all`.
 [profile.default]
 # Surface failing test output inline as it happens, then repeat the
 # summary at the end. On success, stay quiet — use `just test-verbose`
@@ -11,10 +14,8 @@ success-output = "never"
 fail-fast = false
 
 # Network-gated real-world roundtrip overrides. Absorb flaky DNS/HTTP
-# with a couple of retries, warn when a single attempt takes more than
-# 30s (a dead resolver, usually), and hard-kill after 3 minutes to
-# prevent wedged CI. The binary filter covers every `roundtrip_*` test
-# added via the `roundtrip_test!` macro.
+# with retries, warn on slow single attempts, hard-kill wedged runs.
+# Applies whenever the roundtrip binary runs, under any profile.
 #
 # No test-group here: nextest runs each test in its own OS process, so
 # `std::env::set_current_dir` calls are isolated per test. Each of the
diff --git a/Cargo.lock b/Cargo.lock
index 2f277f3..5a1ef16 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -382,6 +382,17 @@ version = "1.0.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b05b61dc5112cbb17e4b6cd61790d9845d13888356391624cbe7e41efeac1e75"
 
+[[package]]
+name = "console"
+version = "0.16.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d64e8af5551369d19cf50138de61f1c42074ab970f74e99be916646777f8fc87"
+dependencies = [
+ "encode_unicode",
+ "libc",
+ "windows-sys 0.61.1",
+]
+
 [[package]]
 name = "core-foundation"
 version = "0.9.4"
@@ -505,6 +516,12 @@ dependencies = [
  "syn 2.0.106",
 ]
 
+[[package]]
+name = "encode_unicode"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "34aa73646ffb006b8f5147f3dc182bd4bcb190227ce861fc4a4844bf8e3cb2c0"
+
 [[package]]
 name = "encoding_rs"
 version = "0.8.35"
@@ -1107,6 +1124,19 @@ version = "2.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f4c7245a08504955605670dbf141fceab975f15ca21570696aebe9d2e71576bd"
 
+[[package]]
+name = "insta"
+version = "1.47.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7b4a6248eb93a4401ed2f37dfe8ea592d3cf05b7cf4f8efa867b6895af7e094e"
+dependencies = [
+ "console",
+ "once_cell",
+ "serde",
+ "similar",
+ "tempfile",
+]
+
 [[package]]
 name = "inventory"
 version = "0.3.21"
@@ -1443,6 +1473,7 @@ name = "mlf-integration-tests"
 version = "0.1.0"
 dependencies = [
  "datatest-stable",
+ "insta",
  "mlf-cli",
  "mlf-codegen",
  "mlf-diagnostics",
@@ -2191,6 +2222,12 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "similar"
+version = "2.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bbbb5d9659141646ae647b42fe094daf6c6192d1620870b449d9557f748b2daa"
+
 [[package]]
 name = "slab"
 version = "0.4.11"
diff --git a/justfile b/justfile
index 3feafcb..dd0aa94 100644
--- a/justfile
+++ b/justfile
@@ -6,18 +6,25 @@
 # Default: run all tests
 default: test
 
-# Install developer tools (test runner, etc.)
+# Install developer tools (test runner, snapshot reviewer)
 install-tools:
     cargo install cargo-nextest --locked
+    cargo install cargo-insta --locked
 
 # Run every test across the workspace (excluding packages we can't build
-# locally). Nextest already parallelises across binaries, so a single
-# invocation is both faster and more comprehensive than chaining the
-# per-suite recipes below — those remain useful for targeted runs.
+# locally). Nextest parallelises across binaries, so a single invocation
+# is both faster and more comprehensive than chaining the per-suite
+# recipes below — those remain useful for targeted runs. Use
+# `just test-fast` to skip network-dependent real-world tests.
 test:
     @echo "Running all workspace tests..."
     cargo nextest run --workspace --exclude tree-sitter-mlf --exclude mlf-wasm
 
+# Run everything except network-dependent real-world roundtrip tests
+test-fast:
+    @echo "Running workspace tests (excluding real-world)..."
+    cargo nextest run --workspace --exclude tree-sitter-mlf --exclude mlf-wasm -E 'not binary(real_world_roundtrip)'
+
 # Run only language tests (mlf-lang crate)
 test-lang:
     @echo "Running mlf-lang integration tests..."
@@ -54,10 +61,14 @@ test-workspace:
     cargo nextest run -p mlf-integration-tests --test workspace_integration
 
 # Run only the network-dependent real-world round-trip tests
-test-real-world:
+test-real:
     @echo "\n🌐 Running real-world round-trip tests (fetches from network)..."
     cargo nextest run -p mlf-integration-tests --test real_world_roundtrip
 
+# Review pending insta snapshot changes (accepts / rejects `.snap.new` files)
+review:
+    cargo insta review
+
 # Run tests with verbose output (stdout streamed live, tests run serially)
 test-verbose:
     cargo nextest run --workspace --exclude tree-sitter-mlf --exclude mlf-wasm --no-capture
diff --git a/tests/Cargo.toml b/tests/Cargo.toml
index f12cb50..34d6d3b 100644
--- a/tests/Cargo.toml
+++ b/tests/Cargo.toml
@@ -20,6 +20,7 @@ toml = "0.8"
 tokio = { version = "1", features = ["full"] }
 tempfile = "3.8"
 datatest-stable = "0.3"
+insta = { version = "1.43", features = ["json"] }
 
 [dev-dependencies]
 # Any additional test dependencies
diff --git a/tests/codegen/lexicon/annotations/expected.json b/tests/codegen/lexicon/annotations/lexicon@annotations.snap
similarity index 73%
rename from tests/codegen/lexicon/annotations/expected.json
rename to tests/codegen/lexicon/annotations/lexicon@annotations.snap
index bc59188..28b674b 100644
--- a/tests/codegen/lexicon/annotations/expected.json
+++ b/tests/codegen/lexicon/annotations/lexicon@annotations.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,7 +12,9 @@
       "key": "custom-key",
       "record": {
         "type": "object",
-        "required": ["text"],
+        "required": [
+          "text"
+        ],
         "properties": {
           "text": {
             "type": "string"
diff --git a/tests/codegen/lexicon/annotations/warnings@annotations.snap b/tests/codegen/lexicon/annotations/warnings@annotations.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/annotations/warnings@annotations.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/basic_record/lexicon@basic_record.snap b/tests/codegen/lexicon/basic_record/lexicon@basic_record.snap
new file mode 100644
index 0000000..fc462b3
--- /dev/null
+++ b/tests/codegen/lexicon/basic_record/lexicon@basic_record.snap
@@ -0,0 +1,30 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
+{
+  "$type": "com.atproto.lexicon.schema",
+  "lexicon": 1,
+  "id": "app.bsky.feed.post",
+  "defs": {
+    "main": {
+      "type": "record",
+      "key": "tid",
+      "record": {
+        "type": "object",
+        "required": [
+          "text",
+          "createdAt"
+        ],
+        "properties": {
+          "text": {
+            "type": "string"
+          },
+          "createdAt": {
+            "type": "string"
+          }
+        }
+      }
+    }
+  }
+}
diff --git a/tests/codegen/lexicon/basic_record/warnings@basic_record.snap b/tests/codegen/lexicon/basic_record/warnings@basic_record.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/basic_record/warnings@basic_record.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/const_float_stringifies/expected.json b/tests/codegen/lexicon/const_float_stringifies/expected.json
deleted file mode 100644
index df16f5c..0000000
--- a/tests/codegen/lexicon/const_float_stringifies/expected.json
+++ /dev/null
@@ -1,20 +0,0 @@
-{
-  "$type": "com.atproto.lexicon.schema",
-  "lexicon": 1,
-  "id": "com.example.const_float_stringifies",
-  "x-threshold": "3.14",
-  "x-revision": 3,
-  "defs": {
-    "main": {
-      "type": "record",
-      "key": "tid",
-      "record": {
-        "type": "object",
-        "required": ["name"],
-        "properties": {
-          "name": {"type": "string"}
-        }
-      }
-    }
-  }
-}
diff --git a/tests/codegen/lexicon/const_float_stringifies/expected_warnings.txt b/tests/codegen/lexicon/const_float_stringifies/expected_warnings.txt
deleted file mode 100644
index 08981b0..0000000
--- a/tests/codegen/lexicon/const_float_stringifies/expected_warnings.txt
+++ /dev/null
@@ -1 +0,0 @@
-com.example.const_float_stringifies: @const("x-threshold", 3.14): ATProto's data model has no floats; emitting as string "3.14" to stay spec-compliant
diff --git a/tests/codegen/lexicon/basic_record/expected.json b/tests/codegen/lexicon/const_float_stringifies/lexicon@const_float_stringifies.snap
similarity index 53%
rename from tests/codegen/lexicon/basic_record/expected.json
rename to tests/codegen/lexicon/const_float_stringifies/lexicon@const_float_stringifies.snap
index 4daf9b2..0ba2d7c 100644
--- a/tests/codegen/lexicon/basic_record/expected.json
+++ b/tests/codegen/lexicon/const_float_stringifies/lexicon@const_float_stringifies.snap
@@ -1,19 +1,24 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
-  "id": "app.bsky.feed.post",
+  "id": "com.example.const_float_stringifies",
+  "x-threshold": "3.14",
+  "x-revision": 3,
   "defs": {
     "main": {
       "type": "record",
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["text", "createdAt"],
+        "required": [
+          "name"
+        ],
         "properties": {
-          "text": {
-            "type": "string"
-          },
-          "createdAt": {
+          "name": {
             "type": "string"
           }
         }
diff --git a/tests/codegen/lexicon/const_float_stringifies/test.toml b/tests/codegen/lexicon/const_float_stringifies/test.toml
deleted file mode 100644
index f63012b..0000000
--- a/tests/codegen/lexicon/const_float_stringifies/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "const_float_stringifies"
-description = "@const with a fractional numeric value stringifies to stay spec-compliant and emits a warning"
-namespace = "com.example.const_float_stringifies"
diff --git a/tests/codegen/lexicon/const_float_stringifies/warnings@const_float_stringifies.snap b/tests/codegen/lexicon/const_float_stringifies/warnings@const_float_stringifies.snap
new file mode 100644
index 0000000..ea2b4dc
--- /dev/null
+++ b/tests/codegen/lexicon/const_float_stringifies/warnings@const_float_stringifies.snap
@@ -0,0 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[
+    "com.example.const_float_stringifies: @const(\"x-threshold\", 3.14): ATProto's data model has no floats; emitting as string \"3.14\" to stay spec-compliant",
+]
diff --git a/tests/codegen/lexicon/doc_comments/expected.json b/tests/codegen/lexicon/doc_comments/lexicon@doc_comments.snap
similarity index 89%
rename from tests/codegen/lexicon/doc_comments/expected.json
rename to tests/codegen/lexicon/doc_comments/lexicon@doc_comments.snap
index 676c8c4..1e962a7 100644
--- a/tests/codegen/lexicon/doc_comments/expected.json
+++ b/tests/codegen/lexicon/doc_comments/lexicon@doc_comments.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
diff --git a/tests/codegen/lexicon/doc_comments/warnings@doc_comments.snap b/tests/codegen/lexicon/doc_comments/warnings@doc_comments.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/doc_comments/warnings@doc_comments.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/implicit_main/expected.json b/tests/codegen/lexicon/implicit_main/expected.json
deleted file mode 100644
index 8389864..0000000
--- a/tests/codegen/lexicon/implicit_main/expected.json
+++ /dev/null
@@ -1,39 +0,0 @@
-{
-  "$type": "com.atproto.lexicon.schema",
-  "lexicon": 1,
-  "id": "com.example.profile",
-  "defs": {
-    "main": {
-      "type": "record",
-      "key": "tid",
-      "record": {
-        "type": "object",
-        "required": ["name"],
-        "properties": {
-          "name": {
-            "type": "string"
-          },
-          "color": {
-            "type": "ref",
-            "ref": "#color"
-          }
-        }
-      }
-    },
-    "color": {
-      "type": "object",
-      "required": ["red", "green", "blue"],
-      "properties": {
-        "red": {
-          "type": "integer"
-        },
-        "green": {
-          "type": "integer"
-        },
-        "blue": {
-          "type": "integer"
-        }
-      }
-    }
-  }
-}
diff --git a/tests/codegen/lexicon/implicit_main/lexicon@implicit_main.snap b/tests/codegen/lexicon/implicit_main/lexicon@implicit_main.snap
new file mode 100644
index 0000000..2214637
--- /dev/null
+++ b/tests/codegen/lexicon/implicit_main/lexicon@implicit_main.snap
@@ -0,0 +1,49 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
+{
+  "$type": "com.atproto.lexicon.schema",
+  "lexicon": 1,
+  "id": "com.example.profile",
+  "defs": {
+    "color": {
+      "type": "object",
+      "required": [
+        "red",
+        "green",
+        "blue"
+      ],
+      "properties": {
+        "red": {
+          "type": "integer"
+        },
+        "green": {
+          "type": "integer"
+        },
+        "blue": {
+          "type": "integer"
+        }
+      }
+    },
+    "main": {
+      "type": "record",
+      "key": "tid",
+      "record": {
+        "type": "object",
+        "required": [
+          "name"
+        ],
+        "properties": {
+          "name": {
+            "type": "string"
+          },
+          "color": {
+            "type": "ref",
+            "ref": "#color"
+          }
+        }
+      }
+    }
+  }
+}
diff --git a/tests/codegen/lexicon/implicit_main/warnings@implicit_main.snap b/tests/codegen/lexicon/implicit_main/warnings@implicit_main.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/implicit_main/warnings@implicit_main.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/integer_constraints/expected.json b/tests/codegen/lexicon/integer_constraints/lexicon@integer_constraints.snap
similarity index 80%
rename from tests/codegen/lexicon/integer_constraints/expected.json
rename to tests/codegen/lexicon/integer_constraints/lexicon@integer_constraints.snap
index eedaccb..56c97bf 100644
--- a/tests/codegen/lexicon/integer_constraints/expected.json
+++ b/tests/codegen/lexicon/integer_constraints/lexicon@integer_constraints.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,7 +12,10 @@
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["age", "temperature"],
+        "required": [
+          "age",
+          "temperature"
+        ],
         "properties": {
           "age": {
             "type": "integer",
diff --git a/tests/codegen/lexicon/integer_constraints/test.toml b/tests/codegen/lexicon/integer_constraints/test.toml
deleted file mode 100644
index 1143de6..0000000
--- a/tests/codegen/lexicon/integer_constraints/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "integer_constraints"
-description = "Test integer_constraints"
-namespace = "com.example.integer_constraints"
diff --git a/tests/codegen/lexicon/integer_constraints/warnings@integer_constraints.snap b/tests/codegen/lexicon/integer_constraints/warnings@integer_constraints.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/integer_constraints/warnings@integer_constraints.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/item_extensions_codegen/expected.json b/tests/codegen/lexicon/item_extensions_codegen/lexicon@item_extensions_codegen.snap
similarity index 64%
rename from tests/codegen/lexicon/item_extensions_codegen/expected.json
rename to tests/codegen/lexicon/item_extensions_codegen/lexicon@item_extensions_codegen.snap
index 8eafe5c..7f91043 100644
--- a/tests/codegen/lexicon/item_extensions_codegen/expected.json
+++ b/tests/codegen/lexicon/item_extensions_codegen/lexicon@item_extensions_codegen.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,9 +12,13 @@
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["name"],
+        "required": [
+          "name"
+        ],
         "properties": {
-          "name": {"type": "string"}
+          "name": {
+            "type": "string"
+          }
         }
       },
       "x-deprecated": true,
diff --git a/tests/codegen/lexicon/item_extensions_codegen/test.toml b/tests/codegen/lexicon/item_extensions_codegen/test.toml
deleted file mode 100644
index 056fab4..0000000
--- a/tests/codegen/lexicon/item_extensions_codegen/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "item_extensions_codegen"
-description = "@const annotations on a record emit as extra JSON fields on the record def"
-namespace = "com.example.item_extensions_codegen"
diff --git a/tests/codegen/lexicon/item_extensions_codegen/warnings@item_extensions_codegen.snap b/tests/codegen/lexicon/item_extensions_codegen/warnings@item_extensions_codegen.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/item_extensions_codegen/warnings@item_extensions_codegen.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/local_references/expected.json b/tests/codegen/lexicon/local_references/lexicon@local_references.snap
similarity index 50%
rename from tests/codegen/lexicon/local_references/expected.json
rename to tests/codegen/lexicon/local_references/lexicon@local_references.snap
index 1bbdbc7..22e30ab 100644
--- a/tests/codegen/lexicon/local_references/expected.json
+++ b/tests/codegen/lexicon/local_references/lexicon@local_references.snap
@@ -1,14 +1,36 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
   "id": "com.example.document",
   "defs": {
+    "metadata": {
+      "type": "object",
+      "required": [
+        "createdAt",
+        "updatedAt"
+      ],
+      "properties": {
+        "createdAt": {
+          "type": "string"
+        },
+        "updatedAt": {
+          "type": "string"
+        }
+      }
+    },
     "main": {
       "type": "record",
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["title", "meta"],
+        "required": [
+          "title",
+          "meta"
+        ],
         "properties": {
           "title": {
             "type": "string"
@@ -19,18 +41,6 @@
           }
         }
       }
-    },
-    "metadata": {
-      "type": "object",
-      "required": ["createdAt", "updatedAt"],
-      "properties": {
-        "createdAt": {
-          "type": "string"
-        },
-        "updatedAt": {
-          "type": "string"
-        }
-      }
     }
   }
 }
diff --git a/tests/codegen/lexicon/local_references/warnings@local_references.snap b/tests/codegen/lexicon/local_references/warnings@local_references.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/local_references/warnings@local_references.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/nested_objects/expected.json b/tests/codegen/lexicon/nested_objects/lexicon@nested_objects.snap
similarity index 75%
rename from tests/codegen/lexicon/nested_objects/expected.json
rename to tests/codegen/lexicon/nested_objects/lexicon@nested_objects.snap
index 33aeafb..d97678f 100644
--- a/tests/codegen/lexicon/nested_objects/expected.json
+++ b/tests/codegen/lexicon/nested_objects/lexicon@nested_objects.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,28 +12,36 @@
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["displayName"],
+        "required": [
+          "displayName"
+        ],
         "properties": {
           "displayName": {
             "type": "string"
           },
           "metadata": {
             "type": "object",
-            "required": ["createdAt"],
+            "required": [
+              "createdAt"
+            ],
             "properties": {
               "createdAt": {
                 "type": "string"
               },
               "settings": {
                 "type": "object",
-                "required": ["theme"],
+                "required": [
+                  "theme"
+                ],
                 "properties": {
                   "theme": {
                     "type": "string"
                   },
                   "notifications": {
                     "type": "object",
-                    "required": ["email"],
+                    "required": [
+                      "email"
+                    ],
                     "properties": {
                       "email": {
                         "type": "boolean"
diff --git a/tests/codegen/lexicon/nested_objects/test.toml b/tests/codegen/lexicon/nested_objects/test.toml
deleted file mode 100644
index a864da0..0000000
--- a/tests/codegen/lexicon/nested_objects/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "nested_objects"
-description = "Test nested_objects"
-namespace = "com.example.nested_objects"
diff --git a/tests/codegen/lexicon/nested_objects/warnings@nested_objects.snap b/tests/codegen/lexicon/nested_objects/warnings@nested_objects.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/nested_objects/warnings@nested_objects.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/record_key_any/expected.json b/tests/codegen/lexicon/record_key_any/lexicon@record_key_any.snap
similarity index 74%
rename from tests/codegen/lexicon/record_key_any/expected.json
rename to tests/codegen/lexicon/record_key_any/lexicon@record_key_any.snap
index d2b37e5..95064fd 100644
--- a/tests/codegen/lexicon/record_key_any/expected.json
+++ b/tests/codegen/lexicon/record_key_any/lexicon@record_key_any.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,7 +12,11 @@
       "key": "any",
       "record": {
         "type": "object",
-        "required": ["title", "content", "slug"],
+        "required": [
+          "title",
+          "content",
+          "slug"
+        ],
         "properties": {
           "title": {
             "type": "string"
diff --git a/tests/codegen/lexicon/record_key_any/test.toml b/tests/codegen/lexicon/record_key_any/test.toml
deleted file mode 100644
index 82771d5..0000000
--- a/tests/codegen/lexicon/record_key_any/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "record_key_any"
-description = "Test record_key_any"
-namespace = "com.example.record_key_any"
diff --git a/tests/codegen/lexicon/record_key_any/warnings@record_key_any.snap b/tests/codegen/lexicon/record_key_any/warnings@record_key_any.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/record_key_any/warnings@record_key_any.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/record_key_literal/expected.json b/tests/codegen/lexicon/record_key_literal/lexicon@record_key_literal.snap
similarity index 78%
rename from tests/codegen/lexicon/record_key_literal/expected.json
rename to tests/codegen/lexicon/record_key_literal/lexicon@record_key_literal.snap
index 716770e..454d968 100644
--- a/tests/codegen/lexicon/record_key_literal/expected.json
+++ b/tests/codegen/lexicon/record_key_literal/lexicon@record_key_literal.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,7 +12,9 @@
       "key": "literal:self",
       "record": {
         "type": "object",
-        "required": ["displayName"],
+        "required": [
+          "displayName"
+        ],
         "properties": {
           "displayName": {
             "type": "string"
diff --git a/tests/codegen/lexicon/record_key_literal/test.toml b/tests/codegen/lexicon/record_key_literal/test.toml
deleted file mode 100644
index 83999c0..0000000
--- a/tests/codegen/lexicon/record_key_literal/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "record_key_literal"
-description = "Test record_key_literal"
-namespace = "com.example.record_key_literal"
diff --git a/tests/codegen/lexicon/record_key_literal/warnings@record_key_literal.snap b/tests/codegen/lexicon/record_key_literal/warnings@record_key_literal.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/record_key_literal/warnings@record_key_literal.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/self_item/expected.json b/tests/codegen/lexicon/self_item/lexicon@self_item.snap
similarity index 64%
rename from tests/codegen/lexicon/self_item/expected.json
rename to tests/codegen/lexicon/self_item/lexicon@self_item.snap
index 6850fb0..d726c83 100644
--- a/tests/codegen/lexicon/self_item/expected.json
+++ b/tests/codegen/lexicon/self_item/lexicon@self_item.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -9,9 +13,13 @@
   "defs": {
     "main": {
       "type": "object",
-      "required": ["value"],
+      "required": [
+        "value"
+      ],
       "properties": {
-        "value": {"type": "string"}
+        "value": {
+          "type": "string"
+        }
       }
     }
   }
diff --git a/tests/codegen/lexicon/self_item/test.toml b/tests/codegen/lexicon/self_item/test.toml
deleted file mode 100644
index 25f768b..0000000
--- a/tests/codegen/lexicon/self_item/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "self_item"
-description = "self {} with docs and @const/@reference extensions emit as top-level JSON fields"
-namespace = "com.example.self_item"
diff --git a/tests/codegen/lexicon/self_item/warnings@self_item.snap b/tests/codegen/lexicon/self_item/warnings@self_item.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/self_item/warnings@self_item.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/string_constraints/expected.json b/tests/codegen/lexicon/string_constraints/lexicon@string_constraints.snap
similarity index 81%
rename from tests/codegen/lexicon/string_constraints/expected.json
rename to tests/codegen/lexicon/string_constraints/lexicon@string_constraints.snap
index a763df2..b04747f 100644
--- a/tests/codegen/lexicon/string_constraints/expected.json
+++ b/tests/codegen/lexicon/string_constraints/lexicon@string_constraints.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,7 +12,10 @@
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["username", "displayName"],
+        "required": [
+          "username",
+          "displayName"
+        ],
         "properties": {
           "username": {
             "type": "string",
diff --git a/tests/codegen/lexicon/string_constraints/test.toml b/tests/codegen/lexicon/string_constraints/test.toml
deleted file mode 100644
index 05404cc..0000000
--- a/tests/codegen/lexicon/string_constraints/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "string_constraints"
-description = "Test string_constraints"
-namespace = "com.example.string_constraints"
diff --git a/tests/codegen/lexicon/string_constraints/warnings@string_constraints.snap b/tests/codegen/lexicon/string_constraints/warnings@string_constraints.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/string_constraints/warnings@string_constraints.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen/lexicon/union_types/expected.json b/tests/codegen/lexicon/union_types/lexicon@union_types.snap
similarity index 51%
rename from tests/codegen/lexicon/union_types/expected.json
rename to tests/codegen/lexicon/union_types/lexicon@union_types.snap
index f627bf7..0304dbf 100644
--- a/tests/codegen/lexicon/union_types/expected.json
+++ b/tests/codegen/lexicon/union_types/lexicon@union_types.snap
@@ -1,3 +1,7 @@
+---
+source: tests/codegen_integration.rs
+expression: output.json
+---
 {
   "$type": "com.atproto.lexicon.schema",
   "lexicon": 1,
@@ -8,21 +12,33 @@
       "key": "tid",
       "record": {
         "type": "object",
-        "required": ["content"],
+        "required": [
+          "content"
+        ],
         "properties": {
           "content": {
             "type": "union",
             "refs": [
-              {"type": "string"},
-              {"type": "integer"},
-              {"type": "boolean"}
+              {
+                "type": "string"
+              },
+              {
+                "type": "integer"
+              },
+              {
+                "type": "boolean"
+              }
             ]
           },
           "media": {
             "type": "union",
             "refs": [
-              {"type": "string"},
-              {"type": "integer"}
+              {
+                "type": "string"
+              },
+              {
+                "type": "integer"
+              }
             ],
             "closed": true
           }
diff --git a/tests/codegen/lexicon/union_types/test.toml b/tests/codegen/lexicon/union_types/test.toml
deleted file mode 100644
index f948118..0000000
--- a/tests/codegen/lexicon/union_types/test.toml
+++ /dev/null
@@ -1,4 +0,0 @@
-[test]
-name = "union_types"
-description = "Test union_types"
-namespace = "com.example.union_types"
diff --git a/tests/codegen/lexicon/union_types/warnings@union_types.snap b/tests/codegen/lexicon/union_types/warnings@union_types.snap
new file mode 100644
index 0000000..23a5161
--- /dev/null
+++ b/tests/codegen/lexicon/union_types/warnings@union_types.snap
@@ -0,0 +1,5 @@
+---
+source: tests/codegen_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/codegen_integration.rs b/tests/codegen_integration.rs
index c332627..9a78f5d 100644
--- a/tests/codegen_integration.rs
+++ b/tests/codegen_integration.rs
@@ -1,31 +1,30 @@
 // Workspace-level integration tests for code generation.
 // Each subdirectory of `codegen/lexicon/` is a test fixture containing
-// `input.mlf`, `expected.json`, and optionally `test.toml`.
+// `input.mlf`, optionally `test.toml`, and insta snapshot files
+// (`lexicon.snap`, `warnings.snap`) checked into the fixture directory.
+//
+// To update snapshots after an intentional change:
+//     just test-codegen         # produces .snap.new files
+//     just review               # interactive accept/reject
 
 use mlf_codegen::generate_lexicon;
 use mlf_integration_tests::test_utils;
 use mlf_lang::{parser::parse_lexicon, Workspace};
-use serde_json::Value;
 use std::fs;
 use std::path::Path;
 
-fn derive_namespace_fallback(test_name: &str) -> String {
-    match test_name {
-        "basic_record" => "app.bsky.feed.post".to_string(),
-        _ => format!("com.example.{}", test_name),
-    }
-}
-
 fn run_lexicon_test(input_path: &Path) -> datatest_stable::Result<()> {
     let test_dir = input_path.parent().ok_or("input.mlf has no parent directory")?;
+    let test_name = test_dir
+        .file_name()
+        .and_then(|s| s.to_str())
+        .ok_or("input.mlf parent has no name")?
+        .to_string();
 
-    let config = test_utils::load_test_config(test_dir, derive_namespace_fallback)
-        .map_err(|e| format!("{}", e))?;
-
-    let namespace = config
-        .test
-        .namespace
-        .ok_or("No namespace specified in test.toml")?;
+    let namespace = test_utils::load_test_config(test_dir, |name| format!("com.example.{}", name))
+        .ok()
+        .and_then(|c| c.test.namespace)
+        .unwrap_or_else(|| format!("com.example.{}", test_name));
 
     let input = fs::read_to_string(input_path)?;
     let lexicon = parse_lexicon(&input).map_err(|e| format!("Failed to parse: {:?}", e))?;
@@ -37,45 +36,22 @@ fn run_lexicon_test(input_path: &Path) -> datatest_stable::Result<()> {
         .map_err(|e| format!("Failed to resolve: {:?}", e))?;
 
     let lexicon = ws.get_lexicon(&namespace).ok_or("Module not found")?;
-
     let output = generate_lexicon(&namespace, lexicon, &ws);
 
-    let expected_str = fs::read_to_string(test_dir.join("expected.json"))?;
-    let expected_json: Value = serde_json::from_str(&expected_str)?;
-
-    if output.json != expected_json {
-        return Err(format!(
-            "Output mismatch:\nExpected:\n{}\n\nGot:\n{}",
-            serde_json::to_string_pretty(&expected_json).unwrap(),
-            serde_json::to_string_pretty(&output.json).unwrap()
-        )
-        .into());
-    }
-
-    let warnings_path = test_dir.join("expected_warnings.txt");
-    let expected_warnings = if warnings_path.exists() {
-        fs::read_to_string(&warnings_path)?
-    } else {
-        String::new()
-    };
-    let actual_warnings = output
+    let formatted_warnings: Vec<String> = output
         .warnings
         .iter()
         .map(|w| format!("{}: {}", w.namespace, w.message))
-        .collect::<Vec<_>>()
-        .join("\n");
-    let actual_warnings = if actual_warnings.is_empty() {
-        String::new()
-    } else {
-        format!("{}\n", actual_warnings)
-    };
-    if actual_warnings != expected_warnings {
-        return Err(format!(
-            "Warnings mismatch:\n--- expected ---\n{}\n--- got ---\n{}",
-            expected_warnings, actual_warnings
-        )
-        .into());
-    }
+        .collect();
+
+    insta::with_settings!({
+        snapshot_path => test_dir.to_path_buf(),
+        prepend_module_to_snapshot => false,
+        snapshot_suffix => test_name,
+    }, {
+        insta::assert_json_snapshot!("lexicon", output.json);
+        insta::assert_debug_snapshot!("warnings", formatted_warnings);
+    });
 
     Ok(())
 }
diff --git a/tests/diagnostics/undefined_reference/errors@undefined_reference.snap b/tests/diagnostics/undefined_reference/errors@undefined_reference.snap
new file mode 100644
index 0000000..fd450fd
--- /dev/null
+++ b/tests/diagnostics/undefined_reference/errors@undefined_reference.snap
@@ -0,0 +1,7 @@
+---
+source: tests/diagnostics_integration.rs
+expression: errors_in_module
+---
+[
+    "UndefinedReference { name: \"User\", span: Span { start: 27, end: 31 }, module_namespace: \"test.post\" }",
+]
diff --git a/tests/diagnostics/undefined_reference/expected.json b/tests/diagnostics/undefined_reference/expected.json
deleted file mode 100644
index 840b6e5..0000000
--- a/tests/diagnostics/undefined_reference/expected.json
+++ /dev/null
@@ -1,9 +0,0 @@
-{
-  "error_count": 1,
-  "errors": [
-    {
-      "code": "mlf::undefined_reference",
-      "message": "UndefinedReference"
-    }
-  ]
-}
diff --git a/tests/diagnostics_integration.rs b/tests/diagnostics_integration.rs
index 6667d1a..edbf3f2 100644
--- a/tests/diagnostics_integration.rs
+++ b/tests/diagnostics_integration.rs
@@ -1,40 +1,26 @@
 // Workspace-level integration tests for diagnostics.
-// Each subdirectory of `diagnostics/` is a fixture with `input.mlf`,
-// `expected.json`, and optionally `test.toml`.
+//
+// Each subdirectory of `diagnostics/` is a fixture with `input.mlf`
+// and `test.toml`. Expected diagnostics are captured as an insta
+// `.snap` file checked into the fixture directory.
+//
+// To update snapshots after an intentional change:
+//     just test-diagnostics    # produces .snap.new files
+//     just review              # interactive accept/reject
 
 use mlf_diagnostics::{get_error_module_namespace_str, ValidationDiagnostic};
 use mlf_integration_tests::test_utils;
 use mlf_lang::{parser::parse_lexicon, Workspace};
-use serde::Deserialize;
 use std::fs;
 use std::path::Path;
 
-#[derive(Debug, Deserialize)]
-struct ExpectedDiagnostic {
-    error_count: usize,
-    errors: Vec<ExpectedError>,
-}
-
-#[derive(Debug, Deserialize)]
-struct ExpectedError {
-    #[allow(dead_code)]
-    code: String,
-    message: String,
-    #[serde(default)]
-    #[allow(dead_code)]
-    span: Option<ExpectedSpan>,
-}
-
-#[derive(Debug, Deserialize)]
-struct ExpectedSpan {
-    #[allow(dead_code)]
-    start: usize,
-    #[allow(dead_code)]
-    end: usize,
-}
-
 fn run_diagnostics_test(input_path: &Path) -> datatest_stable::Result<()> {
     let test_dir = input_path.parent().ok_or("input.mlf has no parent directory")?;
+    let test_name = test_dir
+        .file_name()
+        .and_then(|s| s.to_str())
+        .ok_or("input.mlf parent has no name")?
+        .to_string();
 
     let config = test_utils::load_test_config(test_dir, |name| format!("test.{}", name))
         .map_err(|e| format!("{}", e))?;
@@ -56,8 +42,7 @@ fn run_diagnostics_test(input_path: &Path) -> datatest_stable::Result<()> {
         Err(errors) => errors,
     };
 
-    // Construct the diagnostic to ensure the wiring still works, even if we
-    // don't assert on its rendered form here.
+    // Construct the diagnostic to verify the wiring compiles and runs.
     let _diagnostic = ValidationDiagnostic::new(
         "input.mlf".to_string(),
         input.clone(),
@@ -65,40 +50,20 @@ fn run_diagnostics_test(input_path: &Path) -> datatest_stable::Result<()> {
         validation_errors.clone(),
     );
 
-    let expected_str = fs::read_to_string(test_dir.join("expected.json"))?;
-    let expected: ExpectedDiagnostic = serde_json::from_str(&expected_str)?;
-
-    let errors_in_module: Vec<_> = validation_errors
+    let errors_in_module: Vec<String> = validation_errors
         .errors
         .iter()
         .filter(|e| get_error_module_namespace_str(e) == namespace)
+        .map(|e| format!("{:?}", e))
         .collect();
 
-    if errors_in_module.len() != expected.error_count {
-        return Err(format!(
-            "Expected {} errors but got {}",
-            expected.error_count,
-            errors_in_module.len()
-        )
-        .into());
-    }
-
-    for (i, expected_error) in expected.errors.iter().enumerate() {
-        let actual_error = errors_in_module
-            .get(i)
-            .ok_or_else(|| format!("Expected error #{} but only got {} errors", i + 1, errors_in_module.len()))?;
-
-        let actual_message = format!("{:?}", actual_error);
-        if !actual_message.contains(&expected_error.message) {
-            return Err(format!(
-                "Error #{}: Expected message to contain '{}' but got: {}",
-                i + 1,
-                expected_error.message,
-                actual_message
-            )
-            .into());
-        }
-    }
+    insta::with_settings!({
+        snapshot_path => test_dir.to_path_buf(),
+        prepend_module_to_snapshot => false,
+        snapshot_suffix => test_name,
+    }, {
+        insta::assert_debug_snapshot!("errors", errors_in_module);
+    });
 
     Ok(())
 }
diff --git a/tests/lexicon_to_mlf/all_constraint_keys/expected.mlf b/tests/lexicon_to_mlf/all_constraint_keys/mlf@all_constraint_keys.snap
similarity index 89%
rename from tests/lexicon_to_mlf/all_constraint_keys/expected.mlf
rename to tests/lexicon_to_mlf/all_constraint_keys/mlf@all_constraint_keys.snap
index 1860573..3c7dbec 100644
--- a/tests/lexicon_to_mlf/all_constraint_keys/expected.mlf
+++ b/tests/lexicon_to_mlf/all_constraint_keys/mlf@all_constraint_keys.snap
@@ -1,3 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 def type version = integer constrained {
     minimum: 1,
     maximum: 10,
@@ -35,4 +39,3 @@ def type digest = bytes constrained {
     minLength: 32,
     maxLength: 64,
 };
-
diff --git a/tests/lexicon_to_mlf/all_constraint_keys/warnings@all_constraint_keys.snap b/tests/lexicon_to_mlf/all_constraint_keys/warnings@all_constraint_keys.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/all_constraint_keys/warnings@all_constraint_keys.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/array_item_constraints/expected.mlf b/tests/lexicon_to_mlf/array_item_constraints/mlf@array_item_constraints.snap
similarity index 60%
rename from tests/lexicon_to_mlf/array_item_constraints/expected.mlf
rename to tests/lexicon_to_mlf/array_item_constraints/mlf@array_item_constraints.snap
index 6e17f64..65f6917 100644
--- a/tests/lexicon_to_mlf/array_item_constraints/expected.mlf
+++ b/tests/lexicon_to_mlf/array_item_constraints/mlf@array_item_constraints.snap
@@ -1,7 +1,10 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 def type tags = (string constrained {
     minLength: 1,
     maxLength: 200,
 })[] constrained {
     maxLength: 20,
 };
-
diff --git a/tests/lexicon_to_mlf/array_item_constraints/warnings@array_item_constraints.snap b/tests/lexicon_to_mlf/array_item_constraints/warnings@array_item_constraints.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/array_item_constraints/warnings@array_item_constraints.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/array_union_items/expected.mlf b/tests/lexicon_to_mlf/array_union_items/expected.mlf
deleted file mode 100644
index 7a2a7c4..0000000
--- a/tests/lexicon_to_mlf/array_union_items/expected.mlf
+++ /dev/null
@@ -1,2 +0,0 @@
-def type feed = (post | repost)[];
-
diff --git a/tests/lexicon_to_mlf/array_union_items/mlf@array_union_items.snap b/tests/lexicon_to_mlf/array_union_items/mlf@array_union_items.snap
new file mode 100644
index 0000000..e9153c0
--- /dev/null
+++ b/tests/lexicon_to_mlf/array_union_items/mlf@array_union_items.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
+def type feed = (post | repost)[];
diff --git a/tests/lexicon_to_mlf/array_union_items/warnings@array_union_items.snap b/tests/lexicon_to_mlf/array_union_items/warnings@array_union_items.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/array_union_items/warnings@array_union_items.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/item_extensions/expected.mlf b/tests/lexicon_to_mlf/item_extensions/mlf@item_extensions.snap
similarity index 58%
rename from tests/lexicon_to_mlf/item_extensions/expected.mlf
rename to tests/lexicon_to_mlf/item_extensions/mlf@item_extensions.snap
index d07cc1c..2ce06ec 100644
--- a/tests/lexicon_to_mlf/item_extensions/expected.mlf
+++ b/tests/lexicon_to_mlf/item_extensions/mlf@item_extensions.snap
@@ -1,7 +1,10 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 @main
 @const("x-deprecated", true)
 @const("x-since", "2024-01-01")
 record itemext {
     name!: string,
 }
-
diff --git a/tests/lexicon_to_mlf/item_extensions/warnings@item_extensions.snap b/tests/lexicon_to_mlf/item_extensions/warnings@item_extensions.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/item_extensions/warnings@item_extensions.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/lenient_array_without_items/expected.mlf b/tests/lexicon_to_mlf/lenient_array_without_items/expected.mlf
deleted file mode 100644
index dba5d2f..0000000
--- a/tests/lexicon_to_mlf/lenient_array_without_items/expected.mlf
+++ /dev/null
@@ -1,2 +0,0 @@
-def type tags = unknown[];
-
diff --git a/tests/lexicon_to_mlf/lenient_array_without_items/expected_warnings.txt b/tests/lexicon_to_mlf/lenient_array_without_items/expected_warnings.txt
deleted file mode 100644
index 3540c6e..0000000
--- a/tests/lexicon_to_mlf/lenient_array_without_items/expected_warnings.txt
+++ /dev/null
@@ -1 +0,0 @@
-com.example.arraynoitems: array type is missing `items` field; treating item type as `unknown` (ATProto spec lists `items` as required)
diff --git a/tests/lexicon_to_mlf/lenient_array_without_items/mlf@lenient_array_without_items.snap b/tests/lexicon_to_mlf/lenient_array_without_items/mlf@lenient_array_without_items.snap
new file mode 100644
index 0000000..bc3d06e
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_array_without_items/mlf@lenient_array_without_items.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
+def type tags = unknown[];
diff --git a/tests/lexicon_to_mlf/lenient_array_without_items/warnings@lenient_array_without_items.snap b/tests/lexicon_to_mlf/lenient_array_without_items/warnings@lenient_array_without_items.snap
new file mode 100644
index 0000000..23498e4
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_array_without_items/warnings@lenient_array_without_items.snap
@@ -0,0 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[
+    "com.example.arraynoitems: array type is missing `items` field; treating item type as `unknown` (ATProto spec lists `items` as required)",
+]
diff --git a/tests/lexicon_to_mlf/lenient_empty_union_refs/expected.mlf b/tests/lexicon_to_mlf/lenient_empty_union_refs/expected.mlf
deleted file mode 100644
index 38badcd..0000000
--- a/tests/lexicon_to_mlf/lenient_empty_union_refs/expected.mlf
+++ /dev/null
@@ -1,2 +0,0 @@
-def type items = unknown[];
-
diff --git a/tests/lexicon_to_mlf/lenient_empty_union_refs/expected_warnings.txt b/tests/lexicon_to_mlf/lenient_empty_union_refs/expected_warnings.txt
deleted file mode 100644
index 1696f44..0000000
--- a/tests/lexicon_to_mlf/lenient_empty_union_refs/expected_warnings.txt
+++ /dev/null
@@ -1 +0,0 @@
-com.example.emptyunion: open union has no `refs`; emitting `unknown` as a placeholder (ATProto spec lists `refs` as required on union types)
diff --git a/tests/lexicon_to_mlf/lenient_empty_union_refs/mlf@lenient_empty_union_refs.snap b/tests/lexicon_to_mlf/lenient_empty_union_refs/mlf@lenient_empty_union_refs.snap
new file mode 100644
index 0000000..ca2d220
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_empty_union_refs/mlf@lenient_empty_union_refs.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
+def type items = unknown[];
diff --git a/tests/lexicon_to_mlf/lenient_empty_union_refs/warnings@lenient_empty_union_refs.snap b/tests/lexicon_to_mlf/lenient_empty_union_refs/warnings@lenient_empty_union_refs.snap
new file mode 100644
index 0000000..8dbe9f1
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_empty_union_refs/warnings@lenient_empty_union_refs.snap
@@ -0,0 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[
+    "com.example.emptyunion: open union has no `refs`; emitting `unknown` as a placeholder (ATProto spec lists `refs` as required on union types)",
+]
diff --git a/tests/lexicon_to_mlf/lenient_object_without_properties/expected.mlf b/tests/lexicon_to_mlf/lenient_object_without_properties/expected.mlf
deleted file mode 100644
index 0c4c112..0000000
--- a/tests/lexicon_to_mlf/lenient_object_without_properties/expected.mlf
+++ /dev/null
@@ -1,4 +0,0 @@
-/// Marker for bold text
-def type bold = {
-};
-
diff --git a/tests/lexicon_to_mlf/lenient_object_without_properties/expected_warnings.txt b/tests/lexicon_to_mlf/lenient_object_without_properties/expected_warnings.txt
deleted file mode 100644
index dd12b80..0000000
--- a/tests/lexicon_to_mlf/lenient_object_without_properties/expected_warnings.txt
+++ /dev/null
@@ -1 +0,0 @@
-com.example.marker: object type is missing `properties` field; treating as empty (ATProto spec lists `properties` as required)
diff --git a/tests/lexicon_to_mlf/lenient_object_without_properties/mlf@lenient_object_without_properties.snap b/tests/lexicon_to_mlf/lenient_object_without_properties/mlf@lenient_object_without_properties.snap
new file mode 100644
index 0000000..68bbf11
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_object_without_properties/mlf@lenient_object_without_properties.snap
@@ -0,0 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
+/// Marker for bold text
+def type bold = {
+};
diff --git a/tests/lexicon_to_mlf/lenient_object_without_properties/warnings@lenient_object_without_properties.snap b/tests/lexicon_to_mlf/lenient_object_without_properties/warnings@lenient_object_without_properties.snap
new file mode 100644
index 0000000..d500f8b
--- /dev/null
+++ b/tests/lexicon_to_mlf/lenient_object_without_properties/warnings@lenient_object_without_properties.snap
@@ -0,0 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[
+    "com.example.marker: object type is missing `properties` field; treating as empty (ATProto spec lists `properties` as required)",
+]
diff --git a/tests/lexicon_to_mlf/main_renamed_local_ref/expected.mlf b/tests/lexicon_to_mlf/main_renamed_local_ref/mlf@main_renamed_local_ref.snap
similarity index 56%
rename from tests/lexicon_to_mlf/main_renamed_local_ref/expected.mlf
rename to tests/lexicon_to_mlf/main_renamed_local_ref/mlf@main_renamed_local_ref.snap
index 2875419..074b08d 100644
--- a/tests/lexicon_to_mlf/main_renamed_local_ref/expected.mlf
+++ b/tests/lexicon_to_mlf/main_renamed_local_ref/mlf@main_renamed_local_ref.snap
@@ -1,3 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 @main
 def type thing = {
     value!: string,
@@ -6,4 +10,3 @@ def type thing = {
 def type container = {
     items!: thing[],
 };
-
diff --git a/tests/lexicon_to_mlf/main_renamed_local_ref/warnings@main_renamed_local_ref.snap b/tests/lexicon_to_mlf/main_renamed_local_ref/warnings@main_renamed_local_ref.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/main_renamed_local_ref/warnings@main_renamed_local_ref.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/nullable_fields/expected.mlf b/tests/lexicon_to_mlf/nullable_fields/mlf@nullable_fields.snap
similarity index 74%
rename from tests/lexicon_to_mlf/nullable_fields/expected.mlf
rename to tests/lexicon_to_mlf/nullable_fields/mlf@nullable_fields.snap
index 66a6a71..38f2d8f 100644
--- a/tests/lexicon_to_mlf/nullable_fields/expected.mlf
+++ b/tests/lexicon_to_mlf/nullable_fields/mlf@nullable_fields.snap
@@ -1,3 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 @main
 record profile {
     handle!: string,
@@ -11,4 +15,3 @@ def type settings = {
     theme!: string,
     accent: string | null,
 };
-
diff --git a/tests/lexicon_to_mlf/nullable_fields/warnings@nullable_fields.snap b/tests/lexicon_to_mlf/nullable_fields/warnings@nullable_fields.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/nullable_fields/warnings@nullable_fields.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/ref_hint_warning/expected_warnings.txt b/tests/lexicon_to_mlf/ref_hint_warning/expected_warnings.txt
deleted file mode 100644
index de41d1a..0000000
--- a/tests/lexicon_to_mlf/ref_hint_warning/expected_warnings.txt
+++ /dev/null
@@ -1 +0,0 @@
-com.example.refhint: extension field "xFallback" has value "com.example.other#someType" which looks NSID-shaped; emitted as `@const` — consider `@reference` if you intend workspace name resolution when hand-editing the MLF
diff --git a/tests/lexicon_to_mlf/ref_hint_warning/expected.mlf b/tests/lexicon_to_mlf/ref_hint_warning/mlf@ref_hint_warning.snap
similarity index 56%
rename from tests/lexicon_to_mlf/ref_hint_warning/expected.mlf
rename to tests/lexicon_to_mlf/ref_hint_warning/mlf@ref_hint_warning.snap
index 2c65f0b..f33220c 100644
--- a/tests/lexicon_to_mlf/ref_hint_warning/expected.mlf
+++ b/tests/lexicon_to_mlf/ref_hint_warning/mlf@ref_hint_warning.snap
@@ -1,6 +1,9 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 @main
 @const("xFallback", "com.example.other#someType")
 def type refhint = {
     value: string,
 };
-
diff --git a/tests/lexicon_to_mlf/ref_hint_warning/warnings@ref_hint_warning.snap b/tests/lexicon_to_mlf/ref_hint_warning/warnings@ref_hint_warning.snap
new file mode 100644
index 0000000..3dd0581
--- /dev/null
+++ b/tests/lexicon_to_mlf/ref_hint_warning/warnings@ref_hint_warning.snap
@@ -0,0 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[
+    "com.example.refhint: extension field \"xFallback\" has value \"com.example.other#someType\" which looks NSID-shaped; emitted as `@const` — consider `@reference` if you intend workspace name resolution when hand-editing the MLF",
+]
diff --git a/tests/lexicon_to_mlf/self_top_level_description/expected.mlf b/tests/lexicon_to_mlf/self_top_level_description/mlf@self_top_level_description.snap
similarity index 58%
rename from tests/lexicon_to_mlf/self_top_level_description/expected.mlf
rename to tests/lexicon_to_mlf/self_top_level_description/mlf@self_top_level_description.snap
index 57d6587..98e4829 100644
--- a/tests/lexicon_to_mlf/self_top_level_description/expected.mlf
+++ b/tests/lexicon_to_mlf/self_top_level_description/mlf@self_top_level_description.snap
@@ -1,3 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 /// A short description of the whole lexicon.
 self {}
 
@@ -5,4 +9,3 @@ self {}
 def type described = {
     name!: string,
 };
-
diff --git a/tests/lexicon_to_mlf/self_top_level_description/warnings@self_top_level_description.snap b/tests/lexicon_to_mlf/self_top_level_description/warnings@self_top_level_description.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/self_top_level_description/warnings@self_top_level_description.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/self_top_level_extensions/expected.mlf b/tests/lexicon_to_mlf/self_top_level_extensions/mlf@self_top_level_extensions.snap
similarity index 69%
rename from tests/lexicon_to_mlf/self_top_level_extensions/expected.mlf
rename to tests/lexicon_to_mlf/self_top_level_extensions/mlf@self_top_level_extensions.snap
index b1f1049..8934fc3 100644
--- a/tests/lexicon_to_mlf/self_top_level_extensions/expected.mlf
+++ b/tests/lexicon_to_mlf/self_top_level_extensions/mlf@self_top_level_extensions.snap
@@ -1,3 +1,7 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 /// With extensions.
 @const("revision", 3)
 @const("x-vendor-flag", true)
@@ -8,4 +12,3 @@ self {}
 def type extended = {
     value: string,
 };
-
diff --git a/tests/lexicon_to_mlf/self_top_level_extensions/warnings@self_top_level_extensions.snap b/tests/lexicon_to_mlf/self_top_level_extensions/warnings@self_top_level_extensions.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/self_top_level_extensions/warnings@self_top_level_extensions.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf/unknown_def_type_passthrough/expected.mlf b/tests/lexicon_to_mlf/unknown_def_type_passthrough/mlf@unknown_def_type_passthrough.snap
similarity index 78%
rename from tests/lexicon_to_mlf/unknown_def_type_passthrough/expected.mlf
rename to tests/lexicon_to_mlf/unknown_def_type_passthrough/mlf@unknown_def_type_passthrough.snap
index 98d53be..01aac86 100644
--- a/tests/lexicon_to_mlf/unknown_def_type_passthrough/expected.mlf
+++ b/tests/lexicon_to_mlf/unknown_def_type_passthrough/mlf@unknown_def_type_passthrough.snap
@@ -1,7 +1,10 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: output.mlf
+---
 @main
 @const("type", "permission-set")
 @const("title", "Example Access")
 @const("detail", "Access to example resources.")
 @const("permissions", [{ "type": "permission", "resource": "repo", "collection": ["com.example.one", "com.example.two"] }])
 def type permission = unknown;
-
diff --git a/tests/lexicon_to_mlf/unknown_def_type_passthrough/warnings@unknown_def_type_passthrough.snap b/tests/lexicon_to_mlf/unknown_def_type_passthrough/warnings@unknown_def_type_passthrough.snap
new file mode 100644
index 0000000..87c353b
--- /dev/null
+++ b/tests/lexicon_to_mlf/unknown_def_type_passthrough/warnings@unknown_def_type_passthrough.snap
@@ -0,0 +1,5 @@
+---
+source: tests/lexicon_to_mlf_integration.rs
+expression: formatted_warnings
+---
+[]
diff --git a/tests/lexicon_to_mlf_integration.rs b/tests/lexicon_to_mlf_integration.rs
index 51728c4..4d4ebe3 100644
--- a/tests/lexicon_to_mlf_integration.rs
+++ b/tests/lexicon_to_mlf_integration.rs
@@ -1,11 +1,12 @@
 // Workspace-level integration tests for Lexicon JSON → MLF conversion.
 //
-// Each subdirectory of `lexicon_to_mlf/` is a single test case containing:
-//   - `input.json`: the Lexicon JSON to convert
-//   - `expected.mlf`: the expected MLF source produced by the converter
-//   - `expected_warnings.txt` (optional): the expected warnings, one
-//     per line in `<namespace>: <message>` form. When absent, the test
-//     asserts no warnings were produced.
+// Each subdirectory of `lexicon_to_mlf/` is a test case with an
+// `input.json`. Expected output (MLF source + converter warnings) is
+// captured as insta `.snap` files checked into the fixture directory.
+//
+// To update snapshots after an intentional change:
+//     just test-lexicon-to-mlf  # produces .snap.new files
+//     just review               # interactive accept/reject
 
 use mlf_cli::generate::mlf::generate_mlf_from_json;
 use serde_json::Value;
@@ -14,45 +15,31 @@ use std::path::Path;
 
 fn run_case(input_path: &Path) -> datatest_stable::Result<()> {
     let test_dir = input_path.parent().ok_or("input.json has no parent directory")?;
+    let test_name = test_dir
+        .file_name()
+        .and_then(|s| s.to_str())
+        .ok_or("input.json parent has no name")?
+        .to_string();
 
     let input = fs::read_to_string(input_path)?;
     let json: Value = serde_json::from_str(&input)?;
 
     let output = generate_mlf_from_json(&json).map_err(|e| format!("{:?}", e))?;
 
-    let expected = fs::read_to_string(test_dir.join("expected.mlf"))?;
-    if output.mlf != expected {
-        return Err(format!(
-            "MLF mismatch:\n--- expected ---\n{}\n--- got ---\n{}",
-            expected, output.mlf
-        )
-        .into());
-    }
-
-    let warnings_path = test_dir.join("expected_warnings.txt");
-    let expected_warnings = if warnings_path.exists() {
-        fs::read_to_string(&warnings_path)?
-    } else {
-        String::new()
-    };
-    let actual_warnings = output
+    let formatted_warnings: Vec<String> = output
         .warnings
         .iter()
         .map(|w| format!("{}: {}", w.namespace, w.message))
-        .collect::<Vec<_>>()
-        .join("\n");
-    let actual_warnings = if actual_warnings.is_empty() {
-        String::new()
-    } else {
-        format!("{}\n", actual_warnings)
-    };
-    if actual_warnings != expected_warnings {
-        return Err(format!(
-            "Warnings mismatch:\n--- expected ---\n{}\n--- got ---\n{}",
-            expected_warnings, actual_warnings
-        )
-        .into());
-    }
+        .collect();
+
+    insta::with_settings!({
+        snapshot_path => test_dir.to_path_buf(),
+        prepend_module_to_snapshot => false,
+        snapshot_suffix => test_name,
+    }, {
+        insta::assert_snapshot!("mlf", output.mlf);
+        insta::assert_debug_snapshot!("warnings", formatted_warnings);
+    });
 
     Ok(())
 }