pendpost

Most post schedulers were designed for a world where a person writes the caption, picks the time, and presses send. Pendpost is designed for the opposite case: an AI agent drafts and schedules posts across Instagram, Facebook, LinkedIn, YouTube, X, Telegram, Discord, Mastodon, Nostr, WordPress, and Ghost, while every post sits in a draft queue until a human explicitly approves it or auto-approve is enabled per campaign. The vendor describes the process as ‘agent drafts, you approve, then it ships’ — with the explicit constraint that an agent cannot grant itself autonomy; only the operator can enable auto-approve.

The differentiating feature is the audit trail baked into that gate. Drafting and approving are recorded as separate actor steps, so there is always a traceable answer to ‘who approved this post.’ The caption linter runs against editable brand rules before anything can publish, and the Meta circuit breaker halts outbound sends when the platform starts blocking — behaviors that a queue-and-fire scheduler does not have.

Pendpost fits developers, agencies managing multiple client brands, and technical solopreneurs who want local control over credentials and a clear record of what shipped and who cleared it. It does not fit teams that need a visual drag-and-drop content calendar, non-technical editors who cannot operate a CLI, or organizations that require a hosted multi-user dashboard with role-based permissions. The MIT-licensed local-first core is self-hostable; cloud tiers are paid-only features for teams that want to skip the local setup.

The tool ships with an MCP server alongside the core agent, and the vendor states an API is described in the docs. Setup starts with `npx pendpost`, runs in mock mode by default so credentials are never required to test the full approval loop, and binds exclusively to 127.0.0.1 — nothing phones home from the local install.