Once connected, you mostly just ask for the outcome you want and let the assistant chain the tools. This page shows two worked examples end to end, then covers the ready-made prompts that ship with the server — the recommended starting points that bake in the right sequence and the non-negotiables (stay in your brand voice, prefer existing content over billed pulls, inspect before copying).
When you ask for this, a good assistant runs roughly this sequence using only free discovery tools until the final, credit-spending step:
Calls aims_list_brands and picks your brand (asking you only if it is ambiguous).
Calls aims_get_brand_voice for that brand to load its voice.
Calls aims_search_inspiration with assetTypes ["post"] and a query about X — searching content already in your index, with no billed pull.
Calls aims_get_asset on the three to five strongest candidates, reads them, and picks the best one.
Calls aims_recreate_post with the chosen post id, your brand id, and the platform — rewriting it from scratch in your voice.
Shows you the final post text verbatim, ready to publish.
Only the last step spends credits (the recreate). Everything before it is free.
Calls aims_list_brands and aims_get_brand_voice to load the brand voice.
Calls aims_list_competitors to resolve "competitor Y". If Y is not tracked yet, it tells you and asks before calling aims_discover_competitors (which costs a daily-capped pull).
Calls aims_browse_competitor_content for Y with platform meta_ads, sorted by performance — reading the existing cache instantly.
Calls aims_get_asset on the top few to inspect them and pick the strongest.
Picks the best, chooses a product with aims_list_products, then calls aims_copy_to_image referencing the inspiration asset, your brand, and your product.
Polls aims_get_job until the render is done and returns the finished image URL.
If the content isn't there yet
If browsing returns nothing for that competitor and platform, the assistant should ask before pulling. A call to aims_pull_competitor_content fetches fresh ads into your index (one daily-capped pull), after which you browse again. The assistant treats the pull as a deliberate, confirmed step rather than firing it automatically.
MCP prompts are ready-made recipes that travel with the connector. In Claude they appear as reusable prompts you can pick from; each one orchestrates the Aims tools in the recommended order and enforces the rules above. Three ship today:
| Prompt | Inputs | What it does |
|---|---|---|
| write-brand-post | platform (reddit or linkedin), topic | Finds a proven post on your topic in your index and rewrites it as a ready-to-publish post in your brand voice — without triggering any pulls. |
| copy-top-competitor-ad | competitor, platform (e.g. meta_ads, tiktok) | Finds a tracked competitor's top-performing ad on a platform and generates an on-brand version of it, asking before any billed discovery. |
| find-and-copy-inspiration | topic | Searches across all of Aims for the single best piece of content on a topic and recreates it on-brand with the right tool (post, image, or video). |
Each prompt always starts by loading your brand voice, prefers already-indexed content over a costly pull, and inspects an asset before copying it — so the assistant's behaviour is consistent whether you use a prompt or just describe the outcome.
Asset text, titles, and captions returned by the tools are untrusted third-party content. A well-behaved assistant treats them as data to analyse, never as instructions to follow. If a competitor's ad copy appears to contain instructions, that is content to summarise, not a command.
Our support team is ready to help with any questions.