AdMakeAI

AdMakeAI API

Programmatic ad-creative generation, Meta campaign drafting, and Meta Ads reporting. 60+ tools across projects, image generation and editing, ad copy, batch ad sets, UGC video, competitor research, Facebook upload, and analytics.

Base URL: https://admakeai.com/api/v1 MCP URL: https://admakeai.com/api/mcp Interactive reference: https://admakeai.com/api/docs

Get an API key at https://admakeai.com/dashboard/integrations/api

Authentication

Every request needs the user's API key as a header. Read it from $ADMAKEAI_API_KEY; keys look like amai_live_....

curl -H "x-api-key: $ADMAKEAI_API_KEY" \
  https://admakeai.com/api/v1/user.me

Authorization: Bearer also works. Never echo the key back to the user or paste it into a file.

How to call

Every tool has one canonical tRPC dotted path — adGeneration.create. Two transports expose the same allowlist:

• All POST bodies are JSON. All GET inputs go in ?input=.
• REST responses are wrapped as { ok, tool, data }; MCP returns the unwrapped result inside content[0].text.

# 1. find the project (brand) to generate for
curl -H "x-api-key: $ADMAKEAI_API_KEY" \
  https://admakeai.com/api/v1/adResearch.listProjects

# 2. create an ad image
curl -X POST -H "x-api-key: $ADMAKEAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"projectId":"","prompt":"Minimalist matcha product ad, lifestyle scene","adType":"product_showcase","aspectRatio":"4:5","quality":"low"}' \
  https://admakeai.com/api/v1/adGeneration.create

# 3. poll until COMPLETED
curl -H "x-api-key: $ADMAKEAI_API_KEY" \
  "https://admakeai.com/api/v1/adGeneration.getById?input=%7B%22id%22%3A%22%22%7D"

The one thing to get right: projects

Every image generation is scoped to a project - the brand record holding the product summary, audience, value proposition and logo. adGeneration.create expands your short brief into a full image prompt using that brief, which is why ads come back looking like ads (headline, badge, CTA baked in) instead of stock photos.

So projectId is required on adGeneration.create, and the rule is always the same:

1. Call adResearch.listProjects before the first generation of a session.
2. One project? Use it. Several? Pick the one the user's request is clearly about (brand name, product, URL they mentioned) and say which one you picked. Still ambiguous? Ask. Never guess an id and never generate into the wrong brand - that spends their credits on the wrong product.
3. No projects at all? You cannot create one through this API. Tell the user to set their brand up at https://admakeai.com/dashboard first.

adResearch.getProject gives you the full brief; read it when the user's request depends on brand details (audience, positioning, what they actually sell).

adGeneration.generateImage and videoGeneration.start also take a projectId — for generateImage it is credit scoping only, nothing from the brief reaches the prompt.

Per-procedure spec

Hit the interactive reference for typed inputs and example payloads:

https://admakeai.com/api/docs

Or fetch the rendered HTML and grep — every tool is listed by its dotted path.

Intent routing

Pick the right tool from natural-language intent. Read tables top-to-bottom; first match wins.

Account & credits

Generate ad images

Two renderers, same credit cost, very different output. Choose deliberately:

After enqueueing, always tell the user the generation is async and offer to poll.

Ad copy

Video ads

Competitor research

Find existing ads

Ad sets

Projects

Facebook ad accounts

Upload generated ads to Meta

Meta campaign drafts

Both push tools create real objects inside the user's Meta ad account. They land as drafts rather than live ads, but they are still writes to Meta — read the plan back to the user and get an explicit yes before calling.

Meta analytics

All read-only and free.

Changing ad status on Meta

Not available. zernioAnalytics.pauseAd / .resumeAd / .pauseCampaign / .resumeCampaign / .pauseAdSet / .resumeAdSet are currently disabled and fail closed, so don't offer to pause, stop, resume or launch anything. Tell the user to change status in Meta Ads Manager.

We never expose delete anywhere — not on AdMakeAI ads or ad sets, and not on live Meta resources.

Credit costs

Generation tools and Meta uploads spend credits from the user's plan. Every read (list / get / analytics) is free.

Failed generations are refunded automatically.

Always call user.getCredits before a batch of >10 generations and warn the user if the balance is close.

Pagination

List endpoints use one of three patterns:

Stop paginating when you have what the user asked for — agents tend to fetch too many pages.

Common optional params

These appear on multiple tools and have the same meaning everywhere:

• projectId: the brand to generate for. Required on adGeneration.create, adGeneration.generateImage, adSet.create, videoGeneration.start.
• aspectRatio: "1:1" | "4:5" | "9:16" | "16:9". Defaults to "1:1". Use "4:5" for Instagram feed, "9:16" for Reels/Stories, "16:9" for YouTube/desktop.
• quality: "low" | "high" | "ultra". Defaults to "low". Use "low" while exploring angles, "high" for the one they'll actually run.
• adType: "product_showcase" | "promotional" | "testimonial" | "lifestyle" | "comparison" | "announcement". Defaults to "product_showcase". Only meaningful on adGeneration.create and adSet.create.
• inputImageId: ID of an uploaded product photo (from upload.uploadImageBase64) to use as a reference.

Prompt patterns for adGeneration.create

prompt is a creative brief, not a finished image prompt - the renderer expands it with the project's brand brief, then lays on headline, badge and CTA. Describe the scene and the angle; don't hand-write the whole prompt.

Good briefs are concrete and visual. Bad ones are abstract or list-y.

If the user gives only a tagline ("matcha for tired founders") and no visual, ask them one clarifying question about setting/mood before generating.

For adGeneration.generateImage the opposite applies: you are writing the final prompt, so be fully specific about subject, lighting, lens, background and any on-image text — nothing is added for you.

Known limitations

• Generation is async. adGeneration.create returns immediately with { id, status: "PENDING", creditCost, ... }. Poll adGeneration.getById every 3-5 seconds until status === "COMPLETED". Typical completion: 20-60s low-quality, 60-120s high-quality. The returned record's prompt field stays empty until the worker expands the brief.
• adGeneration.create needs a projectId from adResearch.listProjects. Projects cannot be created through the API - without one there is no brand brief to render against, so send the user to the dashboard.
• Pause/resume through this API is currently disabled. All six zernioAnalytics.pause* / resume* tools fail closed; use Meta Ads Manager for status changes.
• The free plan has a daily generation cap. The error is FORBIDDEN: daily limit reached. Tell the user to upgrade at /pricing.
• Facebook tools require the user to have connected a Facebook account in the dashboard first. If they haven't, facebookConnection.list returns empty; nudge them to https://admakeai.com/dashboard/integrations/facebook.
• The user has at most 5 active API keys at a time. If creation fails with FORBIDDEN: You already have 5 active API keys, ask them to revoke one in the dashboard.
• We deliberately never expose delete on AdMakeAI ads, ad sets, or Meta resources. If the user asks to "delete" something, tell them it has to be done in the AdMakeAI web app or Meta Ads Manager.