flaghand/flaghand

Flaghand is a hosted social publishing desk for AI agents. Your agent (Claude, ChatGPT, Claude Code, Codex, Cursor,
Gemini CLI or your own) drafts posts for X, LinkedIn, TikTok, YouTube and more through Flaghand's MCP server, CLI or
HTTP API. Agents can schedule and manage future posts, with one hourly email showing what changed. You approve
immediate posting with a passkey. Website: https://flaghand.com.
Agents with a shell: sign up over HTTP first (POST https://app.flaghand.com/api/v1/signup). OAuth MCP steps are person-run. Chat apps use the connector.
Agents draft, upload media, check, schedule future posts and request owner approval when needed. Each scheduled post
keeps its exact checked text, media, account and time. Agents never approve, post immediately, start a waiting job or
send a DM for anyone. Request owner approval through Flaghand, not in chat.
## At a glance
What is Flaghand? One hosted MCP server, CLI and HTTP API for drafting, checking and scheduling social posts with
an AI agent, plus a web desk where the owner reviews and approves them. Previews are drawn the way each network shows
the post. Made by Trellisbyte LLC.
Which networks?
X: posts, threads, polls, media and replies.
LinkedIn: personal profiles, not company Pages.
Bluesky: posts and threads.
TikTok: photo and video posts. Until TikTok audits the app, direct posts go only to private accounts; any account
can receive them as drafts in its TikTok inbox (TikTok).
YouTube: videos, Shorts, comment replies and playlist changes (YouTube).
Pinterest: image, carousel and video Pins on business accounts (Pinterest).
How does approval work? Agent tokens can draft, upload media, check drafts, request approval and manage future
scheduled posts without a per-post passkey. Tokens cannot approve or post immediately. For owner approval, the person
sees each post as it will appear and approves its exact text, media, account and time with a passkey (Face ID or Touch ID).
Flaghand posts the exact authorized snapshot; changes need a new schedule or approval. Details: approvals.
What does it cost? One plan: $19 a month or $190 a year (two months free), with 1,500 credits a month. Posting to X
spends credits (3 per post or thread part, 40 for a post with a link); posting to the other networks doesn't. 7 days free
to draft, no card to start. Current prices: https://flaghand.com/pricing, read live from the
plans endpoint.
Which agents work with it? Any MCP client with the server URL https://app.flaghand.com/mcp: Claude and ChatGPT as a connector,
Claude Code, Codex, Cursor and Gemini CLI. Scripts, n8n, Make, Zapier and GitHub Actions use the HTTP API
(https://app.flaghand.com/api/v1, OpenAPI) or the flaghand CLI.
Is it open source? No. This repository holds only the public docs, agent skill and client manifests (MIT-0). The
hosted service, its server and the CLI are private.
## Set up in 3 steps (people)
1. Add the MCP server. In Claude or ChatGPT, add a custom connector with the URL https://app.flaghand.com/mcp. In Claude Code, type
claude mcp add --scope user --transport http flaghand https://app.flaghand.com/mcp, then /mcp. Codex, Cursor and Gemini CLI