diff --git a/CHANGELOG.md b/CHANGELOG.md
index f49c980..56c7005 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- Tag format is now checked on insert. Tags should be no more than 255 characters and match the regex `/\A[\w][\w\-]+[\w]\z/`. [PR #22](https://github.com/riverqueue/riverqueue-ruby/pull/22).
+- Returned jobs now have a `metadata` property. [PR #21](https://github.com/riverqueue/riverqueue-ruby/pull/22).
+
 ## [0.4.0] - 2024-04-28
 
 ### Changed
diff --git a/docs/README.md b/docs/README.md
index 61b354f..c9ca412 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -86,7 +86,7 @@ insert_res.unique_skipped_as_duplicated
 
 ### Custom advisory lock prefix
 
-Unique job insertion takes a Postgres advisory lock to make sure that it's uniqueness check still works even if two conflicting insert operations are occurring in parallel. Postgres advisory locks share a global 64-bit namespace, which is a large enough space that it's unlikely for two advisory locks to ever conflict, but to _guarantee_ that River's advisory locks never interfere with an application's, River can be configured with a 32-bit advisory lock prefix which it will use for all its locks:
+Unique job insertion takes a Postgres advisory lock to make sure that its uniqueness check still works even if two conflicting insert operations are occurring in parallel. Postgres advisory locks share a global 64-bit namespace, which is a large enough space that it's unlikely for two advisory locks to ever conflict, but to _guarantee_ that River's advisory locks never interfere with an application's, River can be configured with a 32-bit advisory lock prefix which it will use for all its locks:
 
 ```ruby
 client = River::Client.new(mock_driver, advisory_lock_prefix: 123456)
diff --git a/driver/riverqueue-activerecord/lib/driver.rb b/driver/riverqueue-activerecord/lib/driver.rb
index 8c329f5..3fce2ee 100644
--- a/driver/riverqueue-activerecord/lib/driver.rb
+++ b/driver/riverqueue-activerecord/lib/driver.rb
@@ -100,6 +100,7 @@ def transaction(&)
         finalized_at: river_job.finalized_at,
         kind: river_job.kind,
         max_attempts: river_job.max_attempts,
+        metadata: river_job.metadata,
         priority: river_job.priority,
         queue: river_job.queue,
         scheduled_at: river_job.scheduled_at,
diff --git a/driver/riverqueue-sequel/lib/driver.rb b/driver/riverqueue-sequel/lib/driver.rb
index 5f615d1..69d99ca 100644
--- a/driver/riverqueue-sequel/lib/driver.rb
+++ b/driver/riverqueue-sequel/lib/driver.rb
@@ -86,6 +86,7 @@ def transaction(&)
         finalized_at: river_job.finalized_at,
         kind: river_job.kind,
         max_attempts: river_job.max_attempts,
+        metadata: river_job.metadata,
         priority: river_job.priority,
         queue: river_job.queue,
         scheduled_at: river_job.scheduled_at,
diff --git a/lib/client.rb b/lib/client.rb
index 1e771ca..42dd5da 100644
--- a/lib/client.rb
+++ b/lib/client.rb
@@ -2,8 +2,13 @@
 require "time"
 
 module River
+  # Default number of maximum attempts for a job.
   MAX_ATTEMPTS_DEFAULT = 25
+
+  # Default priority for a job.
   PRIORITY_DEFAULT = 1
+
+  # Default queue for a job.
   QUEUE_DEFAULT = "default"
 
   # Provides a client for River that inserts jobs. Unlike the Go version of the
@@ -241,7 +246,7 @@ def insert_many(args)
           queue: insert_opts.queue || args_insert_opts.queue || QUEUE_DEFAULT,
           scheduled_at: scheduled_at&.utc, # database defaults to now
           state: scheduled_at ? JOB_STATE_SCHEDULED : JOB_STATE_AVAILABLE,
-          tags: insert_opts.tags || args_insert_opts.tags
+          tags: validate_tags(insert_opts.tags || args_insert_opts.tags)
         ),
         unique_opts
       ]
@@ -260,6 +265,16 @@ def insert_many(args)
     private def uint64_to_int64(int)
       [int].pack("Q").unpack1("q") #: Integer # rubocop:disable Layout/LeadingCommentSpace
     end
+
+    TAG_RE = /\A[\w][\w\-]+[\w]\z/
+    private_constant :TAG_RE
+
+    private def validate_tags(tags)
+      tags&.each do |tag|
+        raise ArgumentError, "tags should be 255 characters or less" if tag.length > 255
+        raise ArgumentError, "tag should match regex #{TAG_RE.inspect}" unless TAG_RE.match(tag)
+      end
+    end
   end
 
   # A single job to insert that's part of an #insert_many batch insert. Unlike
diff --git a/lib/job.rb b/lib/job.rb
index 650fad6..bd3cf0f 100644
--- a/lib/job.rb
+++ b/lib/job.rb
@@ -55,15 +55,15 @@ class JobRow
     # The set of worker IDs that have worked this job. A worker ID differs
     # between different programs, but is shared by all executors within any
     # given one.  (i.e. Different Go processes have different IDs, but IDs are
-    # shared within any given process.) A process generates a new ULID (an
-    # ordered UUID) worker ID when it starts up.
+    # shared within any given process.) A process generates a new ID based on
+    # host and current time when it starts up.
     attr_accessor :attempted_by
 
     # When the job record was created.
     attr_accessor :created_at
 
     # A set of errors that occurred when the job was worked, one for each
-    # attempt.  Ordered from earliest error to the latest error.
+    # attempt. Ordered from earliest error to the latest error.
     attr_accessor :errors
 
     # The time at which the job was "finalized", meaning it was either completed
@@ -79,6 +79,9 @@ class JobRow
     # for the last time and will no longer be worked.
     attr_accessor :max_attempts
 
+    # Arbitrary metadata associated with the job.
+    attr_accessor :metadata
+
     # The priority of the job, with 1 being the highest priority and 4 being the
     # lowest. When fetching available jobs to work, the highest priority jobs
     # will always be fetched before any lower priority jobs are fetched. Note
@@ -112,6 +115,7 @@ def initialize(
       created_at:,
       kind:,
       max_attempts:,
+      metadata:,
       priority:,
       queue:,
       scheduled_at:,
@@ -134,6 +138,7 @@ def initialize(
       self.finalized_at = finalized_at
       self.kind = kind
       self.max_attempts = max_attempts
+      self.metadata = metadata
       self.priority = priority
       self.queue = queue
       self.scheduled_at = scheduled_at
@@ -157,7 +162,7 @@ class AttemptError
     attr_accessor :error
 
     # Contains a stack trace from a job that panicked. The trace is produced by
-    # invoking `debug.Trace()`.
+    # invoking `debug.Trace()` in Go.
     attr_accessor :trace
 
     def initialize(
diff --git a/sig/client.rbs b/sig/client.rbs
index 40f55f0..99a7eae 100644
--- a/sig/client.rbs
+++ b/sig/client.rbs
@@ -16,9 +16,13 @@ module River
     def insert_many: (Array[jobArgs | InsertManyParams]) -> Integer
 
     private def check_unique_job: (Driver::JobInsertParams, UniqueOpts?) { () -> InsertResult } -> InsertResult
-    private def uint64_to_int64: (Integer) -> Integer
     private def make_insert_params: (jobArgs, InsertOpts, ?is_insert_many: bool) -> [Driver::JobInsertParams, UniqueOpts?]
     private def truncate_time: (Time, Integer) -> Time
+    private def uint64_to_int64: (Integer) -> Integer
+
+    TAG_RE: Regexp
+
+    private def validate_tags: (Array[String]?) -> Array[String]?
   end
 
   class InsertManyParams
diff --git a/sig/job.rbs b/sig/job.rbs
index 126f398..fff8c3d 100644
--- a/sig/job.rbs
+++ b/sig/job.rbs
@@ -45,13 +45,14 @@ module River
     attr_accessor finalized_at: Time?
     attr_accessor kind: String
     attr_accessor max_attempts: Integer
+    attr_accessor metadata: Hash[String, untyped]
     attr_accessor priority: Integer
     attr_accessor queue: String
     attr_accessor scheduled_at: Time
     attr_accessor state: jobStateAll
     attr_accessor tags: Array[String]?
 
-    def initialize: (id: Integer, args: Hash[String, untyped], attempt: Integer, ?attempted_at: Time?, ?attempted_by: String?, created_at: Time, ?errors: Array[AttemptError]?, ?finalized_at: Time?, kind: String, max_attempts: Integer, priority: Integer, queue: String, scheduled_at: Time, state: jobStateAll, ?tags: Array[String]?) -> void
+    def initialize: (id: Integer, args: Hash[String, untyped], attempt: Integer, ?attempted_at: Time?, ?attempted_by: String?, created_at: Time, ?errors: Array[AttemptError]?, ?finalized_at: Time?, kind: String, max_attempts: Integer, metadata: Hash[String, untyped], priority: Integer, queue: String, scheduled_at: Time, state: jobStateAll, ?tags: Array[String]?) -> void
   end
 
   class AttemptError
diff --git a/spec/client_spec.rb b/spec/client_spec.rb
index 23ba04a..7b4a7b3 100644
--- a/spec/client_spec.rb
+++ b/spec/client_spec.rb
@@ -51,6 +51,7 @@ def transaction(&)
       finalized_at: nil,
       kind: insert_params.kind,
       max_attempts: insert_params.max_attempts,
+      metadata: nil,
       priority: insert_params.priority,
       queue: insert_params.queue,
       scheduled_at: insert_params.scheduled_at || Time.now, # normally defaults from DB
@@ -194,6 +195,22 @@ def to_json = nil
       end.to raise_error(RuntimeError, "args should return non-nil from `#to_json`")
     end
 
+    it "raises error if tags are too long" do
+      expect do
+        client.insert(SimpleArgs.new(job_num: 1), insert_opts: River::InsertOpts.new(
+          tags: ["a" * 256]
+        ))
+      end.to raise_error(ArgumentError, "tags should be 255 characters or less")
+    end
+
+    it "raises error if tags are misformatted" do
+      expect do
+        client.insert(SimpleArgs.new(job_num: 1), insert_opts: River::InsertOpts.new(
+          tags: ["no,commas,allowed"]
+        ))
+      end.to raise_error(ArgumentError, 'tag should match regex /\A[\w][\w\-]+[\w]\z/')
+    end
+
     def check_bigint_bounds(int)
       raise "lock key shouldn't be larger than Postgres bigint max (9223372036854775807); was: #{int}" if int > 9223372036854775807
       raise "lock key shouldn't be smaller than Postgres bigint min (-9223372036854775808); was: #{int}" if int < -9223372036854775808