Ad Spec Guard

io.github.notformalPublished by GitHub account notformalv1.0.0Updated Nov 22, 2026

Ad platform character limits, creative specs, A/B significance, UTM and policy checks. No account.

The publisher's own server: Tsoden AI Visibility Audit

VerifiedStreamable HTTPWeb executable

Overview

Ad Spec Guard's hosted MCP server (adsbeast.pro); no sign-in required; Streamable HTTP; 9 tools.

No AI overview yet: these facts come from the registry entry and the server itself.

Endpoint
adsbeast.pro
Last checked

Installation

In SourceWeft

  1. Open Ad Spec Guard in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.notformal%2Fad-spec-guard/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "ad-spec-guard": {
      "type": "http",
      "url": "https://adsbeast.pro/api/mcp"
    }
  }
}

README

Ad Spec Guard

Verified specifications and decision thresholds for paid advertising, as a plugin for ChatGPT, Codex and Claude.

No account, no sign-in, no API key.

The problem it solves

Ask any assistant for a Google search headline and you will often get 43 characters. The limit is 30. The copy reads well, nothing in the conversation flags a problem, and the ad is rejected when someone uploads it.

The same failure shows up in other shapes:

  • A 1200x628 image briefed for a vertical story placement.
  • A recommendation to pause a campaign that has collected eleven clicks.
  • An A/B test called after 20 conversions against 25.
  • A budget plan whose conversion rate was an industry average presented as a projection.

None of these are reasoning failures. They are recall failures. The model is remembering a number instead of looking one up, and there is nothing in a chat window that catches it.

This plugin replaces recall with lookup.

Tools

All nine are read-only. They compute from specification tables and stated formulas held in the server, reach no advertising account, and store nothing.

ToolWhat it answers
get_platform_specsThe full specification sheet for a platform: copy limits, image formats, learning-phase windows, bidding prerequisites.
validate_ad_copyWhich drafted fields exceed the platform's hard character limit, and by how much.
validate_creative_specWhether an image or video meets a placement's dimensions, aspect ratio, weight and file type.
check_decision_readinessWhether there is enough data to pause, scale, rotate a creative or switch bidding, and what is still missing.
ab_test_significanceThe p-value of an A/B comparison and the traffic still needed if it is inconclusive.
build_utmA tagged URL with values normalised so they do not fragment into duplicate rows in analytics.
validate_utmWhat is wrong with tagging already on a URL.
estimate_budget_feasibilityWhether a budget can reach a cost or return goal, without filling gaps with industry averages.
check_ad_policy_preflightWhich category and claim patterns in the copy commonly trigger policy rejection.

Platforms covered: Google Ads, Meta, TikTok, LinkedIn, Microsoft Advertising.

What it will not do

  • It will not approve an ad. check_ad_policy_preflight matches patterns and links the governing policy page. Only the platform's review decides whether an ad runs, and clearing the pre-check is not an approval.
  • It will not invent an input. estimate_budget_feasibility never substitutes a benchmark for a figure you did not supply. It solves for the value your goal would require and labels it a condition, not a forecast.
  • It will not answer for a platform it does not hold specs for. An unrecognised platform is an error, not a clean result, because a typo coming back as "no violations found" is worse than no answer.

Install

ChatGPT and Codex

Open Plugins, search for Ad Spec Guard, and select Install. There is no account to connect.

To try it before it is listed, add this repository as a marketplace:

codex plugin marketplace add notformal/adsbeast-plugins

Claude Code

/plugin marketplace add notformal/adsbeast-plugins/plugin install ad-spec-guard@adsbeast-plugins

Any other MCP client

The server speaks streamable HTTP with no authentication:

json
{  "mcpServers": {    "ad-spec-guard": {      "type": "http",      "url": "https://adsbeast.pro/api/mcp"    }  }}

Try it

  • "Write three Google search headlines for a plumbing service and check the limits"
  • "My Meta campaign has 14 clicks and no conversions after 3 days. Pause it?"
  • "Is a 1200x628 image valid for an Instagram story placement?"
  • "Variant A got 20 conversions from 1000 clicks, variant B got 25 from 1000. Is that a real difference?"
  • "Build UTM tags for a Google search campaign called Spring Sale 2026"

Privacy

The server receives the arguments of the tool call and nothing else. It holds no account, sets no cookie, issues no identifier, and writes no request to storage. Requests are not used to train anything.

Do not send personal data in ad copy you are checking. The tools have no use for it, and none of them ask for it.

Full policy: https://adsbeast.pro/privacy

Where the data comes from

Specification tables are maintained by Denha Nexus s.r.o. alongside ADS Beast, an ad operations platform, and are transcribed from the platforms' own published specifications.

Platforms revise limits without notice. Each answer reports its source, and figures that are operating thresholds rather than platform rules say so in the result. Where a number matters for a large build, confirm it against the platform's own specification page.

Support

Issues: https://github.com/notformal/adsbeast-plugins/issues

Licence

MIT. See LICENSE.

Source: ad-spec-guard/README.md at commit 093ff47

Tools

9

get_platform_specs

Get platform specifications

Return the verified specification sheet for one ad platform: hard character limits per copy field, image formats with their required dimensions and file size caps, learning-phase windows, and the conversion history automated bidding needs. Call this before writing ad copy or briefing creative, instead of relying on remembered limits, which change and are frequently misremembered.

Parameterssectionplatformrequired

validate_ad_copy

Validate ad copy against platform limits

Check drafted ad copy against the hard character limits of one platform, field by field, and report which fields exceed the limit and by how much. Use it on every draft before presenting it, because copy over the limit is rejected when the ad is uploaded, not when it is written.

Parametersfieldsrequiredlanguageplatformrequiredcampaign_type

validate_creative_spec

Validate a creative file against platform specs

Check an image or video against a platform placement's required dimensions, aspect ratio, file size cap and accepted file types. Width and height are enough to answer whether a size fits a placement; supply file size and type as well when the file exists, and the result says which checks it could not perform. Returns each violation with its severity and the expected value.

Parameterswidthrequiredheightrequiredplatformrequiredfile_bytesformat_keyrequiredfile_extension

check_decision_readiness

Check whether there is enough data to act

Decide whether a live campaign has produced enough data to justify pausing it, scaling its budget, rotating a creative or switching bidding strategy, and whether it is still inside the platform's learning phase. Returns the unmet requirements with the shortfall and an estimate of how long to wait. Call it before recommending any change and before calling a metric good or bad. For a verdict on a cost per acquisition or return on ad spend use 'judge_metric'; for a verdict on a click-through rate or conversion rate use 'judge_rate_metric', which is gated on clicks rather than conversions.

Parametersspendclicksdecisionrequiredplatformrequiredconversionsimpressionsdays_running

ab_test_significance

Test an A/B result for significance

Run a two-proportion z-test on two variants of an ad, landing page or audience, and report the p-value, whether the difference is statistically significant at the chosen confidence level, and how much more traffic would be needed to resolve it if it is not. Use it for any comparison of two rates instead of judging the gap by eye.

Parametersconfidencevariant_a_trialsrequiredvariant_b_trialsrequiredvariant_a_successesrequiredvariant_b_successesrequired

build_utm

Build a tagged campaign URL

Assemble a campaign URL with consistent UTM tagging, normalise the values so they do not split into duplicate rows in analytics, and report every correction made and every risk found. Also flags a source or medium that departs from the conventional value for the platform, which silently breaks channel grouping.

Parameterstermmediumrequiredsourcerequiredcontentbase_urlrequiredcampaignrequiredplatformconvention

validate_utm

Check tagging on an existing URL

Inspect the UTM parameters already on a URL and report missing required parameters and values that will fragment reporting, such as mixed case or encoded spaces. Does not rewrite the URL.

Parametersurlrequired

estimate_budget_feasibility

Check a budget against a cost or return goal

Work out whether a monthly budget can reach a target cost per acquisition or return on ad spend, and whether it produces enough weekly conversions for automated delivery to leave the learning phase. Nothing is filled in from industry averages: for any input not supplied, it solves for the value the goal would require and labels it as a condition rather than a forecast.

Parametersad_setscurrencytarget_cpatarget_roasexpected_cpcexpected_cvrmonthly_budgetrequiredavg_order_value

check_ad_policy_preflight

Pre-check ad copy for policy risk

Scan ad copy for the two patterns behind most rejections: a business category the platform prohibits or gates behind certification, and a claim the platform treats as unverifiable. Returns each match with the governing policy page. This is a pattern pre-check, not an approval; only the platform's review decides whether an ad runs.

Parameterstextrequiredplatformrequiredverticallanding_url

Version history

1
  1. v1.0.0LatestNov 22, 2026

Score

100% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

Other Tsoden integrations