Arcads Claude Code: Setup, Skills, and Tier Requirements

The arcads-claude-code repository is a third-party open-source skill pack developed by Kruse Media LLC that connects the Arcads video generation engine to terminal-based AI coding agents. At Layer3Labs, we build custom marketing automation workflows and workflow agents for small and mid-sized businesses, and running creative generation from a developer terminal changes how media buyers test creative variants.

The tool allows operators to draft scripts, select direct-to-consumer (D2C) ad layouts, and render user-generated content (UGC) video ads directly from Claude Code or Cursor. It bypasses the manual drag-and-drop web dashboard by routing structured generation commands directly through the Arcads application programming interface (API).

The main hurdle for teams evaluating the repository is account access. Arcads does not publish its API pricing openly, and third-party reviewers report that API keys are restricted to custom Pro plans that cost hundreds of dollars each month. Before cloning the repository, growth teams need to understand what the code executes, how it limits spend, and whether their Arcads subscription can actually authenticate it.

What Arcads Claude Code Is and Who Built It

The arcads-claude-code tool is an open-source integration package published by Kruse Media LLC on GitHub under the permissive MIT license. It holds roughly 1,500 GitHub stars and provides structured agent skills that allow coding assistants to interact with the underlying video models behind Arcads. It is not an official software product created, maintained, or endorsed by Arcads.

Arcads itself is a commercial AI video ad generation platform operated at Arcads. The commercial platform provides a web catalog of more than 1,000 AI actors, video localization across more than 30 languages, and an infinite canvas editor designed for marketing agencies and mobile app studios. While Arcads markets an Arcads AI Video API, Arcads lists no first-party command-line tool or agent skill pack on its site; the pack covered here is community-built.

Kruse Media designed the repository specifically for terminal-driven AI tools such as Claude Code from Anthropic and the Cursor code editor. By packaging model-specific prompts, payload schemas, and template configurations into modular agent files, the repository allows marketers to direct video production through conversational prompts in their terminal.

System Requirements and Installation Steps

Running the skill pack requires a local development environment, an active Arcads account with API credentials, and an AI agent interface. You must install Python version 3.10 or newer on your computer to handle the execution scripts that package ad assets and dispatch network requests.

You also need an active Arcads account that can access the API settings panel at app.arcads.ai/settings/api to generate an API secret key. Third-party reviews report that API keys are only issued on the Pro tier; on a lower plan the settings page will not give you a key to paste into the setup.

Local setup follows a standard open-source workflow. You clone the Kruse Media repository to your machine, run the bundled setup script to install dependencies, assign your Arcads API key to the environment configuration, and open the folder within Claude Code or Cursor.

• Base runtime: Python 3.10 or newer is mandatory for script execution.
• API authentication: a valid Arcads API key from app.arcads.ai/settings/api.
• Agent interface: Claude Code terminal CLI or the Cursor editor environment.
• Optional utility tools: FFmpeg for local video slicing, jq for command-line JSON parsing, Node.js for template helpers, and Whisper for local audio transcription.

Supported Video Models and Meta Ad Templates

The repository organizes its capabilities into a dedicated skills directory containing tailored instruction sets for specific third-party image and video generation models. These skills route parameters to generative engines such as Seedance 2.0, Sora 2 (its API closed September 24, 2026), Veo 3.1, Kling 3.0, Grok Video, and the Nano Banana series, including Nano Banana 2, Nano Banana Pro, and Nano Banana Edit.

For talking-head avatar clips, the repository includes integrations for OmniHuman and audio-driven video pipelines. It also connects to ChatGPT Image 2 for static creative generation and bundles 37 static Meta ad templates.

Workspace context is managed across two core architectural files. MASTER_CONTEXT.md stores persistent workspace memory, recording brand voice guidelines, product value propositions, and historical credit costs so the agent remains consistent across sessions. AGENTS.md carries the multi-assistant guidance, so Claude Code and Cursor read the same rules when either one opens the workspace.

• Generative video skills: dedicated execution files for Seedance 2.0, Sora 2 (discontinued), Veo 3.1, Kling 3.0, Grok Video, and OmniHuman.
• Generative image skills: prompts and payload schemas for Nano Banana 2, Nano Banana Pro, Nano Banana Edit, and ChatGPT Image 2.
• Direct-response templates: 37 static Meta ad templates.
• Workspace memory: MASTER_CONTEXT.md defines persistent brand boundaries, and AGENTS.md coordinates specialized agent behaviors.

Cost Controls and the Credit Logging Loop

Automated video generation requires strict financial safeguards because programmatic loops can rapidly drain an account balance. Arcads bills users per generated video clip, with industry reviewers citing effective costs of approximately $11 for each completed render on standard paid tiers. If an agent loops unchecked while attempting to produce variants, it can exhaust hundreds of dollars in credits in a few minutes.

The arcads-claude-code repository mitigates runaway spending by enforcing a mandatory cost-confirmation checkpoint. Before dispatching any rendering payload to the Arcads endpoints, the agent calculates an expected cost estimate and displays it in the terminal, so you see the expected charge before you tell it to proceed.

To keep its cost estimates accurate over time, the system implements a persistent tracking loop. Every successful generation appends a record to logs/arcads-api.jsonl. The agent reads this log file during subsequent runs to parse real historical creditsCharged values, allowing it to ground future cost estimates in actual billing data rather than static assumptions.

The Reported Pro Plan Requirement for API Access

The biggest obstacle for media buyers adopting arcads-claude-code is that, according to third-party reviews, entry-level Arcads subscriptions do not unlock the necessary API credentials. Arcads does not publish pricing publicly, and visits to arcads.ai/pricing yield an unconfigured 404 error page. Subscription tiers are visible only inside the application dashboard after completing account registration.

Multiple third-party software reviewers, including wireflow.ai and marketermilk.com, report that Arcads divides its catalog into three tiers: Starter at roughly $110 per month for about 10 videos, Creator at approximately $220 per month for 20 videos, and a custom Pro tier cited between $385 and $550 per month. Some reviewers note occasional promotional discounts of roughly 30 percent, bringing entry points down to about $77 or $154 per month.

Those same reviews report that API access and actor cloning sit only on the top Pro tier. A team on the estimated $110 Starter or $220 Creator plan would, on that account, not be able to generate an API key. Using arcads-claude-code therefore starts with a custom Pro plan, before the local code can render a single frame.

Daily Workflow for Paid Social Ad Teams

In a live production environment, arcads-claude-code transforms ad production into an iterative text conversation. The operator initializes the workspace by defining product positioning, target angles, and compliance rules inside MASTER_CONTEXT.md. This file serves as the creative ground truth for the local agent.

Next, the media buyer instructs Claude Code to inspect the 37 Meta ad templates and write five distinct direct-response script angles targeting a single product problem. The agent produces the scripts, selects appropriate AI actor IDs from the Arcads catalog, formats the video generation payloads, and outputs the estimated credit cost.

Once the marketer confirms the cost estimate in the CLI, the agent sends the API requests, records the pending job identifiers, and monitors completion statuses. As finished MP4 clips return, the agent updates logs/arcads-api.jsonl with the exact credits charged, giving the buyer immediate creative assets ready for upload into Meta Ads Manager. In the marketing work we did for MetaKing Studios and HeroMaker Studios, variant volume, not ideas, was the creative-testing bottleneck, which is the problem a terminal loop like this attacks.

Alternatives for Direct Agent Video Generation

Marketers who want automated ad generation without committing to the custom Pro tier have viable alternative routes. If transparent pricing and native AI actors are the primary requirement, Creatify publishes standard self-serve pricing starting at $39 per month for its Starter plan and $99 per month for Pro, which provides access to 1,500 AI actors, product-to-video features (Product Avatar, Product Video, image-to-video), and AI media buying features.

Teams dedicated to programmatic agent workflows can consider Higgsfield, which offers an open Model Context Protocol (MCP) server at Higgsfield MCP. The Higgsfield MCP connector links directly to Claude, Cursor, and ChatGPT without requiring a separate API key, consuming credits from standard self-serve plans that start at $15 per month for 200 credits or $39 per month on an annual Plus subscription. However, Higgsfield focuses primarily on cinematic generation models like Seedance and Kling rather than talking-head UGC actors.

For teams committed specifically to Arcads actors who lack Pro tier budgets, the only practical option is using the standard Arcads web interface on a Starter or Creator plan. While this forfeits the terminal-based speed of Claude Code, it eliminates the expensive custom plan requirement.

• Creatify: self-serve pricing at $39 or $99 monthly with 1,500 actors and product-to-video workflows, detailed on our Creatify vs Arcads comparison.
• Higgsfield MCP: direct agent integration inside Claude and Cursor without custom API tiers, starting at $15 per month, covered in our Higgsfield AI for business guide.
• Arcads Web Dashboard: standard manual creation using the Starter or Creator tiers without API keys, reviewed in our Arcads AI review.
• Detailed pricing mechanics: review historical costs and per-clip estimates on our Arcads pricing breakdown.

Who Should Skip Arcads Claude Code

The arcads-claude-code skill pack is the wrong choice for three distinct groups of marketers. Teams producing fewer than twenty video ad variations per month will not recoup the setup friction and Pro subscription costs, making the web dashboard or cheaper alternatives far more practical.

Performance marketers who are uncomfortable operating inside command-line shells, managing environment variables, or reading JSON logs should also skip this repository. An error in a shell prompt can accidentally trigger unwanted API calls, whereas a web interface provides visual guardrails and confirmation modals.

Finally, bootstrapped direct-to-consumer founders should avoid this tool until they have established profitable creative formulas. Paying an estimated $385 to $550 per month just to access the API creates heavy fixed overhead before proving whether AI avatars convert for their specific audience.

Conditions That Would Change Our Assessment

Our evaluation of arcads-claude-code would change under two clear operational shifts. First, if Arcads opens standard API access on its entry-level Starter plan at $110 per month, the economic barrier to entry would disappear, making this skill pack the cheapest way to automate UGC iteration.

Second, if Arcads releases an official, verified Model Context Protocol server that authenticates using standard browser session tokens, community reliance on third-party repositories would diminish in favor of an officially supported connector.

To test whether terminal-based creative automation fits your performance marketing workflow, check your current Arcads dashboard settings at app.arcads.ai/settings/api to verify whether your account tier permits API key generation before cloning the repository.