Krevaya ships a hosted MCP endpoint, so Claude Code (or Claude Desktop, Cursor, any MCP client) can drive your account — with your permission and inside the same guardrails as the dashboard.
Setup (under two minutes)
1. Mint an API key — Dashboard → Settings → API keys. Scopes:
read— see posts, trends, analyticspublish— create / approve / schedule postsgenerate— run LLM pipelines (repurpose, draft a week)
Give an agent only what it needs — a read-only key makes a great analyst that can't touch your queue.
2. Add the server (the Settings page shows this command pre-filled after minting):
claude mcp add --transport http krevaya https://api.krevaya.com/mcp \
--header "Authorization: Bearer krv_..."
The tools
| Perceive | Act | Generate |
|---|---|---|
| get_analytics — what performed | approve_post / reject_post | repurpose — URL/text → drafts |
| get_audit — the Coach's weekly diagnosis | schedule_post — next free slot or explicit time | draft_week — topic → scheduled week (async) |
| get_queue_summary — what's waiting | create_post — draft or schedule directly | job_status — poll async jobs |
| list_posts, list_trends, get_best_times, whoami | | |
Prompts that work well
- "Check my Krevaya queue — approve the drafts that sound like me, reject anything salesy with a reason, and schedule the approved ones."
- "Look at what performed this month, then draft a week on the theme that's winning."
- "What does my Coach say I should fix this week? Do it if it's something you can do."
Everything an agent does stays inside the product's guardrails: safety screening, plan gates, and review-first semantics apply to API traffic exactly as they do to dashboard clicks.