PostFlow for AI agents

PostFlow is built so an AI agent can do the posting work end to end. It reads what accounts are connected and what each accepts, brings in photos and videos, writes captions, checks every post against each platform’s rules, then saves drafts or schedules them. It can also answer comments and direct messages, and tell you what worked.

Ask your agent things like:

• “Schedule this week’s menu photos to Instagram and Facebook, one a day at 9 am.”
• “Turn this product video into a TikTok and a YouTube Short for Friday evening.”
• “Which posts failed this week, and why? Retry the ones that failed because of a timeout.”
• “Reply to every comment asking about opening hours: we’re open 8 am to 10 pm.”
• “Compare last month’s Reels with photo posts. Which got more reach?”

Pick how your agent connects

Claude and ChatGPTAdd PostFlow as a connector and sign in. No key to copy.

Coding agentsClaude Code, Codex, Cursor, VS Code, Gemini CLI, Windsurf, OpenCode: one command or config block.

Agent SkillFor harnesses that run shell commands but don't speak MCP.

REST API + OpenAPIFor agent frameworks that build tools from an OpenAPI document.

Built to be safe with agents

• Drafts first. Agents are told to save drafts or ask before publishing, unless you asked them to publish.
• Checked before queueing. Captions, media count, file types, sizes, aspect ratios and video length are checked per platform. Problems come back as codes and plain sentences the agent can act on.
• Read-only access. Give an agent a read-only key or connection: it can look and report, but the write tools are hidden.
• Clear tool hints. Tools are marked read-only or destructive, so clients can ask before acting.
• No prompt injection from your inbox. Comments and messages are written by strangers; agents are told never to follow instructions in them, and to show you a reply before sending it.
• Limited and revocable. Credentials can’t connect accounts, invite people or touch billing. Remove them any time in Settings.