moritzkremb

Jev Sales Copilot — Development skill for Claude Code

Development community

Live sales-call copilot on TypeSafe Jev: per-utterance closing probability, signals and next-best-move; syncs to a real recorded call; tailored phrasing via Claude Haiku 4.5, verified by Jev.

How to install Jev Sales Copilot

This entry records only its repository, not the path inside it, so there is no exact command to give. Open moritzkremb/jev-sales-copilot and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Jev Sales Copilot does

Live sales-call copilot on TypeSafe Jev: per-utterance closing probability, signals and next-best-move; syncs to a real recorded call; tailored phrasing via Claude Haiku 4.5, verified by Jev.

Alternatives in Development

Crypto Research Haiku Command — Execute lightweight cryptocurrency research by calling only haiku versions of crypto analysis agents in parall 3.4k ★
MCP Linker — mcp store manager, add & syncs MCP server configurations across clients like Claude code, Cursor💡mcphub 308 ★
Mcptoon — Listing 255 MCP tools costs 71,929 tokens 195 ★

README

Sales Copilot — live closing probability with Jev

A live sales-call copilot. As each utterance of a call arrives (replayed transcript, typed text, or the browser microphone), the app asks TypeSafe's **Jev** (`jev-1.13.0`, a System One model that returns typed decisions instead of text) ~37 atomic questions in **one request**, and within a few hundred milliseconds updates:

Closing probability (0–100 %) as a live line chart — the hero visual.
Live signals — what is going well / badly right now (buying signal, pain identified, budget discussed, decision maker, timeline, objection open, competitor, buyer confusion, rep pitching, rep talking too much, …).
Call stage (opening, discovery, qualification, pitch/demo, objection handling, pricing, closing, next steps).
Open objection type (price, timing, authority, need, trust, competitor) — speculative, confidence-gated.
Next best move — a coaching card picked by Jev from a hand-written 15-move playbook, with the best-fitting pre-written phrasing highlighted. Jev picks; code shows the text. Gated on confidence ("listening…" otherwise).
Talk ratio and pace, computed in code from the transcript.
Tailored phrasing (optional) — when the move changes, a small fast LLM (Claude Haiku 4.5) rewrites the chosen playbook line for this conversation, off the critical path; Jev then judges the sentence and code drops it if it invents a fact. The generic line is always shown first.

Jev never generates text. All arithmetic (composite score, EMA, talk ratio, durations, fact persistence) lives in code. The only generative call is the optional tailored phrasing, and it is gated, verified and never blocking.

Quick start

git clone https://github.com/moritzkremb/jev-sales-copilot.git && cd jev-sales-copilot
cp .env.example .env            # put TYPESAFE_API_KEY=... in it (JEV_API_KEY also accepted)
                                # optional: ANTHROPIC_API_KEY=... enab