Your agent plans the week, writes the posts and schedules them. Nothing goes out until you approve it.
API and MCP on every plan, Free includedApproval on by default
In your AI agent, connected by MCP
YouPlan next week for Northwind: 4 posts. The Ethiopia Guji restock is Thursday morning. Keep X short and put shop links in the first comment.
list_accounts6 accounts
list_posts2 to 8 Nov is empty
create_post4 posts, held for approval
Your agentDone. 4 posts are on your board for 2 to 6 Nov, each waiting for your approval. The X versions fit in 280 characters, and the shop links are in the first comments. Nothing goes out until you approve it.
In PostEveryPlatform
Cold brew, the 18-hour way. The recipe is in the first comment.
Ana's notes from Monday's roast: how she dials in a new bean.
Back in stock: Ethiopia Guji. Peach, black tea and jasmine.
Your weekend brew guide: V60, 15 g, 250 ml, 3 minutes.
Approve sends it on its way. Reject puts it back in drafts.
Your agent works through the same rules you do in the app: your accounts, your plan, your approval.
Add PostEveryPlatform to your agent with the MCP link, an API key, the skill file or the CLI. You choose what it may do.
Say what you want in your own words: plan the week, write a launch post, answer the new comments.
Its posts wait on your board. Approve the ones you like and they go out on time. Reject the rest.
Pick the one your agent understands. Every way uses the same permissions and the same approval, on every plan, Free included.
Works with any MCP client, for example Claude, ChatGPT and Cursor. Add the link, then sign in with your PostEveryPlatform account when your agent asks. No key to paste.
Server link
https://posteveryplatform.com/mcp
Or from the command line
claude mcp add --transport http posteveryplatform https://posteveryplatform.com/mcpMake a key in Settings > AI agents and call the API from anything that speaks HTTP. The OpenAPI file describes every call.
curl https://posteveryplatform.com/api/v1/posts \ -H "Authorization: Bearer $PEP_API_KEY" \ -H "Content-Type: application/json" \ -d '{"text": "Back in stock: Ethiopia Guji", "account_ids": ["acc_threads", "acc_x"], "when": "schedule", "scheduled_at": "2026-11-05T08:00:00Z"}' # The post waits for your approval in the app.
One plain page your agent can read to learn the API: what each call does, which permission it needs and how approval works. Your agent still needs an API key.
https://posteveryplatform.com/skill.md
A small command-line tool for the same API. One line per post, easy to put in a script or a scheduled job.
export PEP_API_KEY=pep_... pep post "Back in stock: Ethiopia Guji" --accounts threads,x --at 2026-11-05T08:00:00Z
Get a signed request at your own https address when a post goes out, partly goes out, fails, or starts waiting for your approval.
post.publishedpost.partialpost.failedpost.pending_approvalAn agent can do a week of work in a minute. You still get the last word on every post.
Posts an agent schedules wait on your board as Holding. You can switch approval off, or let one key publish on its own.
Read, draft, schedule, publish, reply: tick them per key or per connected app. New keys start without publish.
No agent can touch billing, your login, your connected accounts or approvals.
See what each agent did and when. Revoke a key or an app any time and it stops at once.
What this agent may do
Never allowed for any agent: billing, your login, connected accounts, approvals.
Open Agent in the app and chat. It follows the same rules as any agent: your plan, your accounts and your approval. It uses the monthly AI allowance that comes with your plan.
Plan the week, write and schedule posts, and answer comments, in plain words.
YouAny comments worth answering on yesterday's cold brew post?
Standing instructions that run on a schedule. Whatever Autopilot writes waits for your approval, always.
Plan this week's posts from my ideas board.
Held for youWrite a weekend brew guide for X, Threads and Facebook.
Held for youStanding instructions per workspace. Compare the plans
Tap one to copy it, then paste it into your agent or into Agent in the app.
Safety, which agents work, and what it costs.
Setting one up: MCP guide or API reference.
Approval is on by default, so nothing an agent schedules goes out until you approve it. Each agent only gets the permissions you tick, it can never touch billing, your login, your connected accounts or approvals, and every action it takes is in the activity log. You can revoke a key or a connected app at any time.
Any MCP client, for example Claude, ChatGPT and Cursor, through the MCP link. Agents that run commands or read files can use the CLI or the skill file, and anything that speaks HTTP can use the REST API.
No. API keys and MCP are on every plan, Free included. The built-in agent uses the monthly AI allowance that comes with your plan. Autopilot is on Solo (1 standing instruction), Studio (5) and Agency (20).
An agent works inside your plan, just like you. On Free that is 3 accounts and 15 scheduled posts a month; paid plans have unlimited posts. Each key or connected app can make 30 requests a minute on Free and 120 on paid plans.
Yes. Switch it off for the whole workspace in Settings > AI agents, or tick "Posts skip my approval" for one key or one connected app. Posts from Autopilot always wait for your approval.
Settings > AI agents lists every key and connected app, the posts waiting for you, and the activity log. Revoke a key or an app there and it can no longer reach your workspace.
Free plan, no card. API keys and MCP included.