heyreach-jev-bot
Scores HeyReach LinkedIn leads and first-message variants with Jev, picks a message per lead and writes draft campaigns back.
Overview
HeyReach-jev-bot is a Python command-line tool for sales teams that already run LinkedIn outreach through HeyReach. It reads lead lists, campaigns, the sender's network and the inbox, and uses Jev to score each lead against a short description of the offer and ideal customer, score each first-message variant, and estimate the chance a lead replies. Results are written back to HeyReach as tiered lead lists and single-message draft campaigns, with the verdict stored in custom fields such as tier, lead score and reply probability.
The decision model sits between the CRM data and the campaign builder. With 30 or more past conversations it learns from reply history in that account; with none it ranks the 1st-degree network and imported lists against the ICP. The README is explicit that nothing in the tool starts a campaign or sends a message, and it is designed to be driven from Claude Code with a demo mode and a fake-HeyReach test harness.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How well does this lead fit the offer and ideal customer profile?(inferred) | — |
| Yes / no | Will this lead reply to this first message?(inferred) | yes · no |
| Choice | Which first-message variant is best for this person?(inferred) | Variant A · Variant B · Variant C |
How it works
- Describe the offer and target roles in a config file and write 2-3 message variants
- Pull lists, campaigns, network and reply history from HeyReach
- Jev scores lead fit, message variants and reply probability per lead
- Leads are tiered and assigned the best variant; verdicts stored as custom fields
- Dry-run, then create lead lists and DRAFT campaigns; a human launches them
Strengths
- Never sends messages or starts campaigns, so a human reviews every draft
- Demo run on 12 leads is reported to cost about a tenth of a cent
- Per-lead probabilities and reasons are visible inside the HeyReach UI
- Includes a synthetic fake-HeyReach workspace to test judgments without a real account
Caveats
- Requires both a TypeSafe API key and a HeyReach API key
- Learning from history needs about 30 past conversations; cold start relies only on ICP text
- No licence stated in the README
Getting started
Set TYPESAFE_API_KEY and HEYREACH_API_KEY in .env, then run `python -m gtm_bot demo` and later `python -m gtm_bot hr start --config my.json --messages variants.txt`.At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- CLI (designed to be run from Claude Code)
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Signal-based LinkedIn outbound scoring for HeyReach, running on Jev (TypeSafe System One).
logicrw/awesome-jev-projectsSignal-based LinkedIn outbound scoring for HeyReach, running on Jev (TypeSafe System One).
hellogumbo/awesome-jev
Frequently asked questions
What does heyreach-jev-bot decide?
It scores each lead's fit, each message variant, and the likelihood of a reply, then assigns a tier and the best variant per lead.
Will it send LinkedIn messages for me?
No. It only creates lead lists and DRAFT campaigns in HeyReach; you launch them yourself.
What do I need to run it?
Python, a TypeSafe API key for Jev and a HeyReach API key; a demo mode runs on example data first.
Related projects
Self-hostable Reddit buyer-intent finder that has Jev judge every post and comment it reads, with a written reason and cost per lead.
ProjectproductionTypeScript- jev-seo★ 610
Live SEO audit from one homepage URL: crawl, 52 rules, Core Web Vitals and Jev page judgments, reported as PDF, XLSX and Markdown.
PluginbetaPython Rebuilds a site's internal link map with Jev judging thousands of link candidates, raced against Claude Opus 5 on the same queue.
ProjectprototypePythonMarketing and voice-agent platform with an experimental TypeSafe module that picks specialisation labels for agent roles.
ProjectprototypeSite audit for SEO and AI-search visibility: a deterministic crawl, narrow Jev Choice, Score and Noul questions, and a live dashboard.
AppbetaTypeScriptChrome extension that grades readable page sections on ten rubric scores with Jev via Vercel AI Gateway and gives the page an A-E grade.
ProjectbetaTypeScript