Changelog

Connect your own AI agent (MCP server) — Claude Code, Codex CLI, OpenCode, Cursor and VS Code (Copilot) can now connect to your workspace through the CompactHive MCP server — signed in with OAuth or a personal access token. Agents can read app logs, health, builds, deployment history and usage; with opt-in operate access they can also restart, redeploy and configure apps, each change previewed first and audit-logged. Hive Code stays available in the browser for teams without their own agent, now on your own ChatGPT login or OpenAI / Anthropic API key. See Features.

Hive Code — build apps in your browser with AI — Describe what you want and Hive Code's AI agent writes, commits, and ships it — onto your own CompactHive stack (Git, CI/CD, and a managed PostgreSQL), hosted in the EU. No local setup, and everything it produces is a real, exportable Git repo you own. Included on every plan, with bigger burst ceilings on higher tiers. See Features for details.

Stripe live mode — CompactHive can now accept real payments. New tenants get a 14-day no-card trial; after that, monthly subscriptions are billed in EUR. See the SLA for what we commit to and the Terms of Service for billing details.

Complimentary access for friends and pilots — Admins can now mark a tenant as complimentary directly from the management panel. Comp tenants get full platform access, are excluded from billing automation, and never expire. Useful for friend tenants, free pilots, and incident credits.

Per-tenant database dashboards — Internal monitoring now includes per-tenant Postgres metrics — connections, transactions, buffer hit ratio, deadlocks, and PVC usage — backed by kube-state-metrics. Helps us spot problems before customers do.

Public status page — status.compacthive.io is live, hosted on infrastructure independent of the main platform so it survives a primary-site outage. Four monitors cover the platform API, signup flow, marketing site, and the status page itself.

Workflow reliability after Temporal restarts — Fixed a bug where the management app held a stale connection to the workflow engine after server restarts, causing tenant provisioning and deletion to silently fail. Connections are now health-checked and re-established automatically.

Decision Change Protocol — We now follow a written protocol any time we change a customer-facing policy, ADR, or canonical reference. This means stale information across our docs gets cleaned up in the same change — no orphan claims, no silent drift.