Set up a brand from your assistant

You can do everything the Brand tab does without opening RecruitAffiliates.ai. This page lists each setting, the API operation that changes it, and the MCP tool that does the same job. MCP is the standard an assistant such as Claude Code uses to call our tools. The API and the tools run the same code the app runs. A change from your assistant looks exactly like one made on screen.

How a change works

Every tool that changes something answers twice. The first call comes back as a card: one plain sentence saying what will change, what it costs, and a preview_token. Nothing has changed yet.

Your assistant shows you the card. When you agree, it calls again with confirm: true and the token, and the change applies. Change one thing and the token fails.

On the API

The API has no card. Send the change and it applies at once, which is why a key with finder:write should only go to code you trust. Each answer carries an effect, the same sentence the app shows after a save, such as "Saved. Your next weekly search uses these."

What an answer looks like

Every answer from a tool opens with one sentence your assistant can read out. A short list of next steps ends it.

An agent names the brand on every call, so there is no "switch brand" step. Leave it out and the call uses your oldest brand. List your brands with GET /v1/finder/brands or recruit_brands.

Read the settings

Change a setting

Each row is one job. All of these are free: they use no credits.

Put an offer in with structured terms. For 30% for life:

{ "terms": { "rate": { "rate_type": "percent", "rate_bps": 3000 }, "duration": { "kind": "lifetime" } } }

A rate is in basis points: 100 basis points is 1%, so 3000 is 30%.

To preview what strict targeting would hide before you turn it on, use GET /v1/finder/brands/{id}/strict-mode-preview?target_countries=GB,US or recruit_preview_strict_mode. To put back a setting a Not a fit answer changed, use POST /v1/finder/brands/{id}/changes/{change_id}/restore or recruit_restore_change.

Competitors

Add one with a domain and a tier. Direct sells what you sell. Adjacent sells to your buyers. Platform is the affiliate software a rival runs on. You can hold 20 in all: 8 direct, 8 adjacent, 4 platforms.

A change reaches the next weekly search. Nothing is searched or charged on the spot.

People you already work with

These are your exclusions. A person on the list is never offered to you again, and never emailed.

Adding the same person twice is safe. Name the platform for a bare @name; we never guess it. A YouTube @handle counts as unresolved, since looking one up costs money. A Do not contact mark can't be removed this way, on purpose.

Program page, goal and brand

Adding a brand copies the offer, targeting and voice of the brand you copy from, and starts its first search. Your plan sets how many brands you can have. At the limit the call is refused with brand_limit and nothing is written.

First-time setup

One call takes a blank account to a brand with its first search running. Give the domain and country, what it does, its competitors or topics, and the offer.

Send "dry_run": true to the API to see what a setup call would change without changing anything. Setup starts the first search once there is a competitor or topic to look for. The first search uses no credits.

Paid lookups

Two calls spend our money, not your credits. Neither touches your credit balance. Each says the cost first.

• Rival search terms (POST /v1/finder/brands/{id}/rival-terms, recruit_find_rival_terms). Reads each rival's affiliate program page so you can compare terms. With the lookup switched on, it also runs one Google search per rival, which costs up to a fraction of a cent each. The card, or the quote on the API's read, shows the most it can cost. The API refuses to run unless you send confirm_cost_cents at or above that quote.
• Read the site (POST /v1/finder/brands/{id}/research, recruit_research_brand). Uses part of your trial's site-reading allowance, and a brand gets three runs.

Errors