Diagnose your robot pipeline from recorded ROS data.
Merus runs algorithmic diagnostics on ROS bags to map the state of your robot pipeline. Health scores and ranked failure signals give AI agents the evidence to connect symptoms across modules and propose likely fixes tied back to the data.
pip install merus-rossoonA real walking trial with command dropout and joint-effort spikes.
Unitree G1 data from a recorded ROS bag. Real recording, real findings: command dropout, joint-effort spikes, and timing irregularities. The numbers below come from running the Merus worker locally.
The pipeline
From a rosbag drop to deploy-blocking evidence.
Five stages, each anchored to a real surface in the product. No mockups.
Ingest
Upload a .mcap, .mcap.zstd, .db3, or .bag file. Cloud uploads stay under the hosted size ceiling while local runs can process larger bags when RAM allows.
bag.mcap · 412 MB · 1.38M msgs · 276s
Diagnose
Diagnostic modules run across timing, TF, QoS, sensors, navigation, and communication evidence. Numeric SLIs use a geometric mean; warning and critical findings apply an ordinal severity veto.
signals: timing + control + sensors · score: 0.499 (critical)
Explain
Structured diagnostic context flows into the shipped AI report path. Eight tool calls max, then ranked findings, evidence, and a fix plan.
AI review · bounded tools · evidence-linked report
Inspect
Open the diagnostic view to inspect ranked findings, module metrics, plots, and optional 3D viewer artifacts generated during upload.
plots: 18 · viewer.rrd: optional · findings: 12
Export
Export diagnostic JSON or the AI report Markdown for review notes. The raw evidence stays visible next to the recommendation.
merus-analysis.json · merus-report.md
The diagnostics
0 algorithms. 6 categories. Supported ROS bags.
Auto-selected per bag. Geometric-mean aggregation — one critical signal collapses the score by design, matching ROS diagnostic_aggregator policy.
Pipeline
- · jitter
- · gaps
- · bandwidth
- · qos_mismatch
- · rosout
Sensor
- · imu_allan
- · camera_freshness
- · lidar_density
- · battery_drop
Signal
- · changepoints
- · anomalies
- · allan_variance
- · value_changepoints
Navigation
- · stuck_detection
- · odom_consistency
- · path_tracking
- · slam_convergence
Control
- · cmd_vel
- · joint_state_health
- · joint_limits
- · trajectory_tracking
TF
- · timing_analysis
- · rotation_stability
Sample diagnostic
The robot on the left. The diagnostic on the right.
Open source
The engine is yours. The convenience is ours.
Good diagnostics should be inspectable. Hosted upload, AI report generation, and assisted Team onboarding are the paid convenience layer.
AGPL worker. Public package next.
The diagnostic engine lives in the worker tree under AGPL-3.0. The packaged CLI is coming next so teams can run it behind their firewall.
pip install merus-rossoon- Audit every module
- All 37 modules in the worker tree
- Run locally
- Zero raw data leaves your machine
- Extend checks
- Contribution path planned
Hosted convenience.
Drop a bag, get a ranked AI report. Keep diagnostic history. Export evidence. Skip the infrastructure.
- →Hosted parser · no env to configure
- →AI likely-cause reports · Claude Sonnet 4
- →Diagnostic history · optional 3D scene playback
- →JSON/Markdown exports · evidence preserved
Pricing
Pricing for real diagnostic work
Start free. Add AI when you ship. Scale when you're not alone.
For engineers validating diagnostics on their first rosbags.
- 5 diagnostics per month
- Current Merus diagnostic catalog
- Metrics, plots, and ranked findings
- ROS2 MCAP/DB3 + partial ROS1 bag support
- 1 robot
- 30-day diagnostic retention
- AI likely-cause reports
For engineers debugging field data and shipping robots.
- 50 diagnostics per month
- AI likely-cause reports
- Current Merus diagnostic catalog
- Diagnostic metrics, plots, and exports
- 1 robot
- 90-day diagnostic retention
- Read API access
- Email support within 48h
For robotics teams piloting shared diagnostics with assisted onboarding.
- 200 pooled diagnostics per month
- 3 seats with manual workspace onboarding
- 3 robots
- AI likely-cause reports
- 1-year diagnostic retention
- Diagnostic exports for review
- Self-serve team invitations
For OEMs and fleets that need scoped deployment and support.
- Everything in Pro
- Scoped diagnostic and robot quotas
- Private worker or on-prem deployment scoping
- Security and SSO planning support
- Custom ROS/runtime integration scoping
- Custom retention scoped to deployment
- Pilot support with dedicated contact
Get started
Turn a rosbag into an evidence-backed diagnostic report.
Free to start. Sample bags included. No credit card.
37 modules · .mcap · .db3 · .bag · 90s on 100 MB