Inspection got easier.
MCAP is easier to inspect, validate, and move between tools than a lot of older recording setups. That is why it keeps showing up in robotics stacks.
MCAP to point cloud
MCAP to point cloud, in your browser.
MCAP is becoming the default recording format in modern robotics workflows because it is portable, indexed, and easier to work with than the old pile of bag formats. But the tooling gap is still real: inspect the recording, yes; turn it into a mapped point cloud, not always.
Why engineers moved to MCAP
Better recording format, same mapping problem.
Mapping did not.
Once you have the file, you still need to answer the hard question: do you need raw frames from one topic, or do you need a mapped output from the full run?
Three upstream ways to inspect MCAP
Good tools for inspection, not a complete map pipeline.
MCAP CLI
Start by summarizing the file and confirming the topics you care about.
mcap info demo.mcap mcap cat demo.mcap --topics /velodyne_points
Foxglove
Foxglove is excellent for opening the file and visually inspecting the data stream.
foxglove-studio /path/to/your/file.mcap
Great for playback and debugging. It is not a SLAM-to-map service.
Python
If you want to inspect or filter ROS1 messages inside MCAP, the upstream Python helpers get you there quickly.
from mcap_ros1.reader import read_ros1_messages
for msg in read_ros1_messages("demo.mcap"):
print(f"{msg.topic}: {msg.ros_msg}")Inspection is not mapping
When you need a mapped output, hand the MCAP to LidarFlow.
What the mapping path looks like
Inspect, validate, reconstruct, review.
$mcap info demo.mcap
$topics: 12 messages: 184k
$duration: 00:04:21
$/velodyne_points · PointCloud2
mcap info
Start by confirming the topics, message types, and duration are what you expect.
Validate · Topics
Topic suggestions
LidarFlow scans the MCAP and proposes the LiDAR, GNSS, IMU, and TF topics for the run.
Georef · WGS84
SLAM + georeferencing
GLIM runs the SLAM step. When GNSS is available FlexCloud aligns the result to real-world coordinates.
SLAM · Trajectory
lowhigh
Mapped point cloud
Preview the mapped output in the browser, then download PCD, PLY, and metadata.
The limit of the upstream path
You can inspect the recording and still not have a map.
The CLI, Foxglove, and Python libraries are exactly what you want when the job is “tell me what is inside this file.” They are not the full answer when the job is “hand me back one mapped point cloud from the run.”
Where LidarFlow fits
It closes the last mile from recording to mapped artifact.
Upload the MCAP, let the product validate the topics, run LiDAR SLAM, georeference the result when GNSS exists, and download the mapped artifacts from the browser.
Related guides