Documentation/Deploy & connect
DEPLOY & CONNECT

Connect only what you need.

Choose manual control or approved automation, and prepare your workspace to keep publishing while you are away.

6 min read

Personal reminders with Telegram

Telegram is optional. Create your own bot, configure its token and webhook secret, and connect the owner’s private chat using the kit’s connection instructions.

Use notification preferences to choose personal reminders, post reminders, reply reminders, and assistant updates. Quiet hours and one shared daily notification limit apply across assistant reminders and reply automation notices, using your workspace timezone.

Quiet hours silence notifications; they do not pause approved X publication or Reply Target collection. Use the separate automation controls when you want to pause that work.

The reminder integration is for the owner’s private Telegram chat. Group chats are not the supported setup.

Optional X API publishing

Manual copy-and-publish works without X developer credentials. For API publishing, bring an eligible X app and your own API access. Set its callback URL to your application’s /api/x/callback route.

Configure X_CREDENTIAL_KEY as a base64-encoded 32-byte value, then connect and test from Settings. Publication workflows start disabled and require explicit approval and the appropriate access grants.

Switch to the intended account profile before connecting X. For original posts and threads, approve the exact saved content and connected destination, then choose Publish now or a future API time. A manual reminder or a new MCP key does not authorize publication.

Generate X_CREDENTIAL_KEY
node -e "console.log(require('node:crypto').randomBytes(32).toString('base64'))"

This key uses base64, unlike the example hex setup secrets. X API eligibility and charges are separate from this kit.

Understand what accepting a reply target permits

Optional collection gathers eligible reply targets from configured sources. Accept a target in the app or your private Telegram chat to authorize one generated reply to that target. A separately running, permitted Codex or Claude client can then generate and publish the wording without another review.

Keep manual reply preparation if you want to inspect every reply before sending it yourself. Collection, publication, replies, and self-replies have independent switches. Approval may expire, and accepted work waits when no agent is available.

Review the X connection's capability checks and platform requirements before enabling automation. Owner permission does not replace the access, eligibility, or approvals required by X.

Going on vacation? Keep your presence going.

Prepare your posts before you leave, then let your hosted workspace deliver the versions you have approved. This uses the existing publication, reply, and assistant controls; there is no separate global vacation switch.

  1. Finish the initial deployment and owner setup. Check your host's scheduler: Cloudflare uses its scheduled Worker, VPS Compose includes a scheduler, and Vercel or Netlify needs an external authenticated call every minute. Keep that hosting and scheduling available while you are away.
  2. Switch to the intended X account profile, connect your eligible X API account, and test the connection. Enable API publication and check the relevant permissions and automation controls.
  3. Prepare and save your posts or threads. Review the exact text, media, and destination account, approve API publication, and choose explicit future delivery times. A calendar plan or manual reminder alone does not send a post. Your hosted scheduler can deliver these approved originals without an AI client staying online.
  4. Choose how you want to handle replies. Keep manual mode to review suggestions and send replies yourself. For automatic replies, enable the relevant controls, grant your assistant Allow approved publication, and accept specific eligible targets in the app or Telegram. Each accepted target permits one generated reply without another wording review, subject to expiry and the configured limits.
  5. If you want AI preparation or approved automatic replies during your trip, connect the included private MCP server to a compatible client and configure that client's schedule separately. Keep its runner available. Local scheduling needs the relevant computer and client available; hosted client execution needs its own working setup. Saving a routine in OwnPost does not start the client.
  6. Check a scheduled delivery and any assistant workflow before you leave. Set optional Telegram notifications, quiet hours, and limits to suit your trip. Review delivery that needs attention; check X before retrying an uncertain result. On your return, review the published posts and record their results.

New reply targets still need your acceptance, and accepted targets can expire while waiting for an agent. Quiet hours affect notifications, not approved post delivery. The kit is a one-time purchase; ongoing hosting, AI, and X API costs, connection maintenance, updates, and backups remain part of running your own workspace.

Routines and your AI client

The app tracks routines and due work. The configured scheduler handles reminders and approved jobs, but does not launch an AI model. Cloudflare uses a scheduled Worker; Node deployments use authenticated cron calls.

If you want scheduled AI preparation, configure a schedule in your compatible client separately. Saving a routine in the workspace does not create that client’s automation.

Your workspace should feel like yours. Start small, then make it fit.