Tools· Marketing & Campaigns

    claude-ads

    Paid media audit suite for Claude Code across twelve advertising platforms (Google, Meta, TikTok, LinkedIn, Microsoft, YouTube and more), built from skills and audit agents, read-only by default, open source.

    reviewinganalyticalexecutableOpen Source
    Source code
    Recommended: /plugin marketplace add AgriciDaniel/claude-ads, or after cloning, bash install.sh --source=local (Windows: install.ps1 -Source local). No manual copying of individual files.
    claude-ads

    Description

    Strengths

    Separate scoring axes
    Health, evidence coverage, regulatory exposure, and opportunities are reported separately instead of being merged into a single score.
    Honestly flagged incomplete runs
    If a required worker is missing, the run is marked incomplete rather than silently presented as a complete audit.
    Six-step approval gate for account changes
    Applying a change needs a tested capability, explicit IDs, a before/after diff with blast radius, owner approval, an idempotency key with rollback, and a final state verification.
    Consistent structure across platforms
    Each of the twelve advertising platforms gets its own skill, audit worker, control reference, and capability declaration, built to the same pattern.
    One rule set, several hosts
    Claude Code is the intended runtime, and according to the project the same skill files can also run under Codex, Gemini, Cursor, Windsurf, and Goose where the host supports it.

    Assessment

    AI features

    • Audit with dated evidence Evaluates ad accounts and returns findings with a date and explicit confidence instead of a blanket assessment.
    • Campaign and budget planning Plans channels, budgets, measurement, and experiments for paid media campaigns.
    • Creative briefs and assets Produces briefs and assets for copy, image, video, and product photos.
    • Ongoing monitoring Tracks pacing, delivery, tracking, ad fatigue, policy compliance, and performance.
    • Versioned reports The canonical result is validated JSON, from which Markdown, HTML, and optionally PDF are rendered.
    • Drafted account changes Drafts safe changes to ad accounts but does not apply them by default.

    Suitable for

    • Agencies and inhouse performance teams who want to audit ad accounts consistently across multiple platforms
    • Teams who want to produce campaign creatives in text, image, and video from skills in a structured way rather than manually
    • People who need an explicit approval and rollback process before any account change
    • Less suitable for Less suitable for immediate, automated live optimization without human sign-off; the system is deliberately conservative

    Limitations and notes

    • No live writes without the full approval chain Without a tested capability, explicit IDs, a diff, approval, an idempotency key, and verification, every change stays blocked; missing ceilings always mean no write.
    • Permanent deletion not supported According to the project, permanently deleting objects is not supported in v2.
    • Extra system dependencies for browser and PDF Screen capture requires an operator-installed Playwright browser payload, PDF output requires the system libraries WeasyPrint and Pango.
    • Limited Python support The managed dependencies are built for CPython 3.11 and 3.12 on the documented Linux, macOS, and Windows wheel matrix; other interpreters are rejected before any change is made.
    • Two distribution paths with different access The public version at AgriciDaniel/claude-ads is free under the MIT license, a community mirror with earlier access runs through the paid AI Marketing Hub Pro membership.

    Quick start

    1. Install via the native plugin flow: /plugin marketplace add AgriciDaniel/claude-ads, then /plugin install claude-ads@ai-marketing-hub-claude-ads.
    2. Alternatively clone locally with git clone https://github.com/AgriciDaniel/claude-ads.git and run bash install.sh --source=local, on Windows .\install.ps1 -Source local.
    3. Store credentials for the desired advertising platforms in environment variables, an OS keychain, or an approved secret manager, not in the repository.
    4. For screen capture, install a Playwright browser payload; for PDF export, provide the system libraries WeasyPrint and Pango.
    5. Start a first audit via the platform-specific command, for example /ads google on a standalone install or /claude-ads:ads google as a Claude Code plugin.

    Tips

    • Before every run, check that credentials sit in environment variables or a secret manager and not in the repository or in reports.
    • On a run flagged as incomplete, check which worker failed before passing the finding along as a complete audit.
    • Keep an eye on evidence coverage in the report; values below 60 percent count as insufficient evidence, 60 to 79 percent only as provisional.
    • For a skill-only install without managed dependencies, run the installer with --no-deps or -NoDeps.
    • Before removing the tool, use bash uninstall.sh --target=claude or uninstall.ps1 so only the manifest-owned files get deleted.

    Access

    Last reviewed: · Pricing, plans and features are a snapshot in time. Check the provider's own page before deciding.

    In the workshop this becomes your method.

    Whoever sees this process run once wants the agent behind it next. We build that in the workshop From Process to Agent.

    View workshops

    Related resources

    Browse all resources

    Conversation, not pitch

    Understand first, then decide. We take time for an initial conversation, without sales pressure, without obligation.

    Schedule a call