How to make videos with Make

1. Scenario 1, scheduled Weekly (Monday 9:00): a Make AI Agents (New) agent with the Cliptude server added through Add MCP (list_templates and estimate_credits only) and a research tool you choose writes the week's narration and returns the title, script, template_id and credits.
2. MCP Client → Call a tool: create_job with the template_id, title, script and a thumbnail_reference_url. Cliptude creates nothing yet and returns the review (settings, warnings, credits) and a review_token, valid for 24 hours.
3. Data store: save the arguments and the review_token under one key. Slack or Email: send yourself the review in plain words with an Approve link to a custom webhook URL that carries that key.
4. Scenario 2, started by the custom webhook: read the record, then Call a tool → create_job again with the same arguments plus review_token (and acknowledge, if the review listed items you accept). Save the returned job id.
5. Scenario 3, scheduled every 15 minutes: Call a tool → get_job for each saved job id. When the status is completed, post the video link to Slack and delete the record. Before you publish, open the job in Cliptude, or ask a chat agent to run open_editor_project and view_scenes, and look at every frame.

Source: Make Apps: MCP Client (open beta)

Make has no skills folder and the skills CLI has no Make target. The nearest thing is the agent's system prompt, which Make describes as where you "define the agent's purpose and constraints"; it applies in every scenario that uses the agent, and each run can add Additional system instructions mapped from scenario data.

Paste the rules from the cliptude-script skill (research the topic, write the spoken narration, never pass the title as the prompt) into the system prompt, and the parts of the cliptude skill about templates and the review. The cliptude-plan skill directs every shot of an agent-planned video and is over 70 KB; that work suits a coding agent such as Claude Code better than a scheduled scenario.

Keep the tools that spend credits out of the agent. Turn on list_templates, estimate_credits and get_job for it, and call create_job from a Call a tool module after your approval, so a scenario never starts a video on its own.

You plan long-form 16:9 YouTube videos with the Cliptude tools.
- Never pass the video title as prompt. Research the topic, write the spoken narration, and return it as script.
- Call list_templates, then estimate_credits with the template and script.
- Return the title, script, template_id and credits. Do not start a job.

https://github.com/cliptude/cliptude/tree/main/mcp_server/skills/cliptude-script
https://github.com/cliptude/cliptude/tree/main/mcp_server/skills/cliptude

Source: Make Help: Make AI Agents app (Run an agent, system prompt)

All three can call Cliptude's MCP server from a schedule. Make suits you if your automations already live there and you like a visual canvas with per-action credits. Zapier has the simplest built-in approval step but counts every completed action as a task. n8n can run on your own server and bills per workflow run, so a long polling loop costs less there.

Source: Make pricing and credits FAQ

Make bills in credits. The Free plan has 1,000 credits a month; paid plans (Core, Pro, Teams, Enterprise) add more, and annual billing is cheaper than monthly. Every module action uses credits, most of them one; Router and error handler modules use none. Unused credits expire at the end of the term.

A weekly video scenario uses a few credits per run, plus one per get_job check and, if you use Make's AI Provider for the agent or Execute an action with AI, credits for AI tokens. You can instead connect your own OpenAI account and pay OpenAI. The video itself is charged to your Cliptude credits, on a plan from $19; Cliptude has no free plan.

Source: Make pricing and credits FAQ

Make (make.com) is a visual automation platform. You build scenarios from modules for more than 1,000 apps, run them on a schedule or from a webhook, and pay in credits: most module actions use one credit. Its MCP Client app, in open beta, connects scenarios and AI agents to remote MCP servers such as Cliptude.

Call a tool or Execute an action with AI?

Call a tool runs the Cliptude tool you name with the inputs you map, so it suits fixed steps such as create_job and get_job. Execute an action with AI hands the server's tools to a model that picks them for a task you describe; Make suggests testing with Call a tool first.

Make's MCP server or the MCP Client app?

Make also runs its own MCP server, which lets AI assistants run your Make scenarios. The MCP Client app is the other direction: it lets Make call an outside server such as Cliptude.