As of March 2026, the main costs in this package come from Claude API usage and any relay hosting you choose to add.
When you see token pricing, think of it as paying for how much text the system reads and writes.
| Role or task | Model | Why |
|---|---|---|
| Planning Lead | Claude Sonnet 4.6 | This is one of the higher-judgment jobs in the system, so it uses the stronger model. |
| Briefing Lead | Claude Sonnet 4.6 | Briefings are stakeholder-facing, so the higher-quality model is used. |
| Slack and channel replies | Claude Sonnet 4.6 | Replies are user-facing and benefit from better reasoning and writing quality. |
| Research Analyst | Claude Haiku 4.5 | This work is more about extraction and summarization than deep reasoning. |
| Editorial Director | Claude Haiku 4.5 | This role writes short review notes, so the lighter model is enough. |
| Knowledge Manager | Claude Haiku 4.5 | This role produces short cleanup recommendations rather than long reasoning chains. |
| Program Manager | Claude Haiku 4.5 | This role produces concise task-queue guidance. |
| Channel context extraction | Claude Haiku 4.5 | This is intake filtering, so the lighter model keeps cost down. |
| Cost area | What to expect | Notes |
|---|---|---|
| Claude Haiku 4.5 | $1 per 1M input tokens, $5 per 1M output tokens | Used for lighter extraction, filtering, and review work. |
| Claude Sonnet 4.6 | $3 per 1M input tokens, $15 per 1M output tokens | Used for higher-judgment planning, briefings, and user-facing replies. |
| Cloudflare Workers relay hosting | Free plan available; paid plan starts at $5/month, then usage charges beyond included limits | Official published overage examples include about $0.30 per 1M requests and $0.02 per 1M CPU ms on paid usage. |
| Vercel relay hosting | Hobby is free; Pro starts at $20/month and includes $20 usage credit | Best if you already use Vercel. Good for Node-based relay examples. |
| WhatsApp messaging | Variable | Meta pricing depends on country, message type, and who initiated the conversation. |
| Usage style | What it looks like | Cost profile |
|---|---|---|
| Lighter daily rhythm | GitHub + Gmail only, default office hours, a few manual runs, no Slack yet. | Usually the lightest ongoing cost profile. |
| Typical working rhythm | GitHub + Gmail + Slack, default office hours, regular briefings, moderate messaging. | A reasonable middle ground for a small team. |
| Heavier operating rhythm | Many enabled sources, frequent manual runs, lots of messaging, large documents, richer briefings. | This is where spend rises more noticeably. |
The default times are a starter template, not a rule.
Intake Lead: as often as useful updates actually arrivePlanning Lead: once each morningDelivery Lead: once later in the dayBriefing Lead: once a weekIf your team starts later, change the times. What matters most is choosing a rhythm that matches how your team actually works.