Velocity
Velocity is Claude Code for Product Managers — an agentic AI assistant that connects to your real PM tools and helps you make better product decisions.
Core trick
An orchestrator delegates to four subagents that pull Linear, Slack, and Amplitude data, then cites sources in generated PRDs.
Description
Velocity is Claude Code for Product Managers — an agentic AI assistant that connects to your real PM tools and helps you make better product decisions. Powered by Claude Opus 4.6 as an orchestrator with 4 specialized subagents (Research, Backlog, Prioritization, Doc Writer), it pulls live data from Linear, Slack, Amplitude, and Notion, then synthesizes it into actionable outputs. One prompt can trigger 5+ tool calls across 4 external services — research Slack conversations, pull engagement metrics, write a data-grounded PRD, publish it as a GitHub Gist, create a tracked Linear issue, and notify the team. Every claim is grounded in real data with source citations. Each session runs in an isolated Daytona sandbox for security. Built entirely solo during the hackathon with Claude Opus 4.6 and the Claude Agent SDK.
README excerpt
# Velocity — Claude Code for PMs
Claude Code for Product Managers — an agentic AI assistant that connects to your real PM tools and helps you make better product decisions. Pulls live data from Linear, Slack, and Amplitude, then synthesizes it into actionable outputs: PRDs, Linear issues, Slack updates — all grounded in real data.
Built with Claude Opus 4.6 and the Claude Agent SDK for the Anthropic Build with Claude Hackathon (Feb 2026).
## What It Does
**Ask Velocity anything a PM would ask their team:**
- "What should we prioritize this sprint?" — pulls Linear issues + Slack discussions, runs RICE scoring
- "Mark VEL-25 as done" — updates Linear tickets directly
- "Write a PRD for the new onboarding flow" — generates grounded documents citing real data
- "What are customers saying about search?" — searches Slack conversations for feedback patterns
- "Compare options A vs B" — applies weighted scoring frameworks with evidence
Velocity doesn't just chat — it **acts**. It reads and writes to your tools, cites sources for every claim, and coordinates specialized subagents to handle complex multi-step workflows.
## Architecture
```
User → Next.js UI → FastAPI (SSE bridge) → Claude Agent SDK (Opus 4.6)
│
┌─────────┼─────────┐
│ │ │
Research Backlog Prioritization