> ## Documentation Index > Fetch the complete documentation index at: https://docs.adsmcp.si/llms.txt > Use this file to discover all available pages before exploring further. # Diagnostic & Helper Tools > ADSMCP's free, read-only get-started, pre-flight, and diagnostic tools — validate a campaign draft, explain platform errors, and confirm a campaign is live. Quota-free on every plan. ADSMCP ships a group of **9 read-only "get unstuck" tools** that help you start, pre-flight, and troubleshoot — without touching a single campaign. They cover the moments around the work: figuring out the best next step, checking a campaign draft against the platform's rules before you create it, turning a cryptic platform error into a plain-language fix, and confirming a campaign actually went live. > [!NOTE] > > **These tools are quota-free.** Calling any Diagnostic & Helper tool never counts against your monthly tool-call limit, and they're available on **every plan, including Free**. They're also all **Read** tools — they never create, edit, or pause anything. Any actual fix is something your assistant does afterward with the normal platform tools, with your confirmation. Your AI assistant discovers and calls these automatically — you don't need to memorize names. Ask things like *"what should I do first?"*, *"will this campaign draft be accepted?"*, *"why did that fail?"*, or *"is my campaign actually running?"* and the right tool runs. ## Getting started The fastest way to go from "connected" to "doing something useful." Each tool reads your real connected accounts and data, so the suggestions name your business and platforms — not generic advice. | Tool | What it does | | - | - | | `start_here` | The single best next step for your exact state — not connected yet, connected but no data, or already running campaigns — with a worked example prompt. | | `suggest_next_action` | Ranks the highest-value things you can do right now from your live data: analyze, find wasted spend, set a monitor, optimize, or launch. | | `list_what_i_can_do` | A personalized capability menu organized by goal (Analyze & Report, Monitor & Alert, Manage & Optimize, Create & Launch), filtered to your connected platforms. | Personalized first step based on your connected accounts. I just connected ADSMCP — what should I do first? ## Before launching a campaign Catch the problems that cause "campaign creation failed" *before* anything reaches the ad platform. These work across Google Ads, Meta, TikTok, and LinkedIn. | Tool | What it does | | - | - | | `get_campaign_spec` | Returns the exact required fields, character limits, asset requirements, and allowed values for a given campaign type, so a create satisfies every constraint. | | `validate_campaign_draft` | Dry-run validates a fully assembled campaign draft against the platform's spec (no write) and returns per-field issues — character limits, item counts, missing required fields. | Check a campaign draft against the platform's rules before creating it. Before you create anything, show me the spec for a Meta conversions campaign and dry-run my draft against it. ## When something fails Turn an opaque ad-platform error into a clear cause and next step — and confirm a fix actually worked. | Tool | What it does | | - | - | | `explain_platform_error` | Translates a raw ad-platform error message into a plain-language cause and the exact step to fix it. | | `why_did_this_fail` | Summarizes your recent failed actions and the errors behind them, so the issue can be corrected and retried. | | `verify_campaign_is_live` | After you create a campaign, checks it actually exists and is serving (not paused, not on a \$0 budget) with a **real live check against the platform** — Google Ads, Meta, TikTok, or LinkedIn. | > [!TIP] > > `why_did_this_fail` is the one to reach for when you're not sure *where* a problem is. It names the real error behind each of your recent failed calls, which is usually enough to tell an expired **AI-client connection** (your ChatGPT/Claude link to ADSMCP) apart from an **ad-platform sign-in** — two different connections that customers regularly confuse. See [Authentication](https://docs.adsmcp.si/knowledge-base/security.md) for how the two layers differ, and [Support](https://docs.adsmcp.si/knowledge-base/support.md) for the reconnect steps. Find the real error behind a failed action. My Google Ads data isn't loading. Show me why my recent calls failed and what the actual error was. ## See your value What your plan has actually delivered this period. | Tool | What it does | | - | - | | `usage_value_summary` | Your plan as value delivered: tool calls used vs. included this period and what those calls accomplished. | For "what moved since last week", ask for performance directly — your assistant reads it from the platform tools on [Google Ads](https://docs.adsmcp.si/ad-platforms/google-ads.md), [Meta](https://docs.adsmcp.si/ad-platforms/meta-ads.md), [TikTok](https://docs.adsmcp.si/ad-platforms/tiktok-ads.md), [LinkedIn](https://docs.adsmcp.si/ad-platforms/linkedin-ads.md), or [Amazon](https://docs.adsmcp.si/ad-platforms/amazon-ads.md), which cover a date range you choose rather than a fixed window. See what your plan delivered this period, then what moved. Summarize what my ADSMCP usage delivered this period, then show me what changed across my ad accounts in the last 7 days. ## Interactive cards in ChatGPT & Claude In **ChatGPT** and **Claude**, these tools render as branded, interactive cards instead of plain text — a setup checklist, a usage meter, an account switcher. The cards adapt to light and dark themes, and the buttons can kick off the next step. On any client that doesn't render widgets (and in terminal clients like [Claude Code](https://docs.adsmcp.si/ai-clients/claude-code.md) or [Cursor](https://docs.adsmcp.si/ai-clients/cursor.md)), the same answer falls back to clean text — nothing is lost, it just isn't a card. > [!NOTE] > > Interactive cards are built on the open **MCP-Apps** standard (SEP-1865) that ChatGPT and Claude both support. They render entirely from the tool's own data — no external page loads — and any link (upgrade, connect a platform) opens through your client to [adsmcp.ai](https://adsmcp.si). ## Availability | | | | - | - | | **Plans** | All, including **Free** | | **Quota** | Quota-free — never counts against your tool-call limit | | **Tool type** | Read-only (no campaign changes) | | **Clients** | Every connected AI client; interactive cards render in ChatGPT and Claude | | **Platforms** | Google Ads, Meta, TikTok, LinkedIn | ## Recently withdrawn Four tools were pulled from this group on **August 24, 2026**: `diagnose_my_setup`, `preflight_campaign`, `whats_changed_since_last_visit`, and `weekly_opportunities`. They misread which ad platform you meant — saying "meta" could fail to match the account's canonical `meta_ads` name and produce a "no account connected" answer on an account that was connected and working. Your assistant no longer sees them, so treat them as gone rather than waiting for them to come back. The nine tools above are unaffected. For the job `diagnose_my_setup` used to do, ask *"why did that fail?"* — see the [Tip](#when-something-fails) above. ## Related Documentation * [Tool Catalog](https://docs.adsmcp.si/agent-skills/tools.md) — every ADSMCP tool by platform * [Agent Skills Overview](https://docs.adsmcp.si/agent-skills/overview.md) — how skills enforce the right workflow * [Authentication & Security](https://docs.adsmcp.si/knowledge-base/security.md) — the difference between your AI-client connection and ad-platform sign-in * [Multi-Account](https://docs.adsmcp.si/knowledge-base/multi-account.md) — switching active accounts from chat * [Support](https://docs.adsmcp.si/knowledge-base/support.md) — when to escalate beyond self-diagnosis * [Prompt Engineering Playbook](https://docs.adsmcp.si/agent-skills/prompt-engineering-playbook.md) — A custom skill, account dossier, and prompt book generated from your own ad account (\$39 one-time, one free on annual) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.