@slatesvideo/cli

The slates command for the Slates AI video studio. Drive Slates from your terminal, or let Codex (with your ChatGPT account) or Claude Code shell out to it instead of loading 92 tool schemas into context.

Install

npm i -g @slatesvideo/cli

Requires Node.js 18+ and the Slates desktop app (slates.video).

Commands

| Command | What it does | |---|---| | slates login | Authorize this machine via magic link (or --token slates_sk_...) | | slates logout | Clear the stored cloud token | | slates status | Show connection state, account, and credit balance | | slates mcp | Detect installed MCP clients and print the exact config for each; --write merges it into Claude Desktop / Cursor configs (with a .bak backup) | | slates install-skills | Install the bundled agent skills into .claude/skills/<name>/SKILL.md; --global targets ~/.claude/skills | | slates run <op> | Invoke any Slates operation by id; --list shows all 92, <op> --help shows one op's flags | | slates run <op> --input '<json>' | Pass a NESTED object (Shot params/refs, batch updates) — flags cannot express those | | slates use <project> | Set the default project, by id or name, so ops stop needing --projectId | | slates doctor | Check every setup precondition and print the fix for each failure | | slates completion bash\|zsh\|pwsh | Shell completion for the operation ids |

Skills

slates install-skills writes each one to .claude/skills/<skill-name>/SKILL.md in your current project, which is the layout Claude Code's skill discovery requires. Use --global to install for every project. Restart Claude Code afterward, then ask: "what slates skills do you have?" to verify.

Using slates run (agent-driven)

# List every operation
slates run --list

# Create a project
slates run slates_create_project --name "neon samurai"

# Estimate before generating (the cost-discipline skill makes agents do this)
slates run slates_estimate_generation_cost --model kling-v3-standard-5s

# Generate an image into a project
slates run slates_generate_image --projectId <uuid> --prompt "..." --resolution 1k --aspectRatio 16:9

# Kick off a video in the background, then poll it
slates run slates_generate_video --projectId <uuid> --model veo-3.1-fast --prompt "..." --aspectRatio 16:9 --duration 8 --background true --confirm true
slates run slates_get_generation_status --generationId <uuid>

# Assemble + export
slates run slates_add_clip_to_timeline --projectId <uuid> --assetId <uuid>
slates run slates_export_video --projectId <uuid> --outputPath "C:\\Videos\\ad.mp4"

# Structured output for scripting
slates run slates_get_credit_balance --json

The timeline, export, background-generation, edit-image, and image-reference ops need a Slates desktop on agent API v2 — if an op reports a version error, update Slates (Settings → Check for Updates) and retry.

Notes for agent use:

• Repeated flags or comma lists become arrays: --ids a,b,c.
• --json emits {text, data, images}. Image entries carry mimeType and byte count only, not the binary. Use the MCP server (@slatesvideo/mcp-server) when the agent needs to see generated images inline.
• Generation ops gate on missing aspectRatio/resolution and on cost over 17 credits (confirm=true required), so a scripted run cannot silently overspend. The exit code says which: 0 done, 1 error, 3 confirm required, 4 clarification required, 5 desktop unreachable, 6 not signed in.

First-time setup

1. Install and open the Slates desktop app.
2. Settings → Agent Control → enter your email → Send link. Click the emailed link.
3. slates status should now show your account. (Alternative: slates login.)

Credentials live in ~/.slates/agent-connection.json. No env vars.

Links

• Website: slates.video
• Source: github.com/EricDisero/slates-mcp
• Issues: github.com/EricDisero/slates-mcp/issues