Early access
Pose CLI — motion analysis for serious work.
A command-line and MCP toolkit that turns 3D pose data into reproducible motion reports. Built for sports scientists, PE programs, performance analysts, and the teams who need to ship findings — not slide decks.
$ pose analyze data/session.csv \ --tool detect_velocity → 38 swings detected → peak velocity 28.4 m/s → stability index 0.72 $ pose chat ./data → Launching Claude Code with MCP tools…
Built for the people who study motion
Three groups we hear from most.
Sports scientists
Run reproducible analysis across sessions and athletes, export markdown / JSON for papers and reports.
PE programs
Standardized motion statistics for student cohorts — quality flags, joint amplitude, swing counts.
Performance analysts
Sport-specific rule evaluation (badminton today, more sports next), batch processing, CI-friendly JSON output.
Capabilities
Pluggable analysis tools with a consistent interface.
Motion statistics
Frame rate, jitter ratio, joint amplitude, data-quality assessment.
Swing detection
Multi-feature: velocity peaks, shoulder engagement, direction consistency, arm-involvement filtering.
Energy analysis
Peak energy detection and velocity profiling with configurable time windows.
Sport-specific rules
Badminton rule compliance today, more sports planned. Bring-your-own rule sets supported.
AI chat (MCP)
pose chat launches Claude Code with all analysis tools pre-loaded for interactive exploration.
Gap registry
Crowdsourced capability requests; temporary implementations auto-generated for unsupported analyses.
Architecture
A familiar Python stack built around pluggable MCP tools.
- FastAPI REST API for batch and CI workflows
- MCP Server with pluggable analysis tools
- PostgreSQL for session metadata, feedback, gap registry
- Docker on Railway (staging + prod), local dev on
localhost:8000 - Single JointData3D format shared with PoseCoach
Data format
Bring your own pose data — Pose CLI works with our compact 13-column CSV.
timestamp, joint, wx, wy, wz, mx, my, mz, ldx, ldy, ldz, oreason, jitter
Binary .bin from PoseCoach is also accepted natively. Output as Markdown reports or CI-friendly JSON.
Ready to put your motion data to work?
Pose CLI is in early access while we onboard partners. Drop us a line and tell us what you want to study.