Product API for developers

JSON over HTTPS

One endpoint to send products and tags, a few to read status, limits and tagged products. No SDK required.

1,000 per request

Batches are validated, queued and applied in order by a worker.

Safe retries

An idempotency key makes a repeated request return the same batch instead of duplicating work.

Per-item errors

One bad product doesn't fail the batch — it is reported with its id and reason.

Getting started

Create a Product API source in the panel (Sources → Add source → Product API). Each source has its own id; an account can have several.
Create an API key on the API Access page. The key is shown once and stored only as a hash; it can be limited to certain IP addresses and given an expiry date.
Send your products to the batch endpoint with the key as a bearer token.
Poll the batch until it is applied, and read any per-item errors.

Keys are only accepted over HTTPS and never in a query string. The API host sets no cookies and answers JSON only.

OpenAPI description

curl -X POST https://api.feedmanager.org/v1/sources/42/products/batch \
  -H "Authorization: Bearer $FEEDMANAGER_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: 2026-09-20-0001" \
  -d '{
        "operation": "upsert",
        "products": [
          {"id": "SKU-1042", "title": "Trail Runner GTX",
           "price": "74.90 EUR", "availability": "in_stock",
           "link": "https://store.example/p/1042"}
        ]
      }'
import os, httpx

key = os.environ["FEEDMANAGER_KEY"]
products = [{"id": "SKU-1042", "title": "Trail Runner GTX",
             "price": "74.90 EUR", "availability": "in_stock"}]

response = httpx.post(
    "https://api.feedmanager.org/v1/sources/42/products/batch",
    headers={"Authorization": f"Bearer {key}",
             "Idempotency-Key": "2026-09-20-0001"},
    json={"operation": "upsert", "products": products},
    timeout=30,
)
batch = response.json()
print(batch["batch_id"], batch["status"])
const response = await fetch(
  "https://api.feedmanager.org/v1/sources/42/products/batch",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.FEEDMANAGER_KEY}`,
      "Content-Type": "application/json",
      "Idempotency-Key": "2026-09-20-0001",
    },
    body: JSON.stringify({
      operation: "upsert",
      products: [{ id: "SKU-1042", title: "Trail Runner GTX",
                   price: "74.90 EUR", availability: "in_stock" }],
    }),
  },
);
const batch = await response.json();
console.log(batch.batch_id, batch.status);

Reference

Endpoints

Base address https://api.feedmanager.org/v1. Every response is JSON and is never cached.

Product API endpoints
Endpoint	What it does
POST `/sources/{id}/products/batch`	Send up to 1,000 products. `operation` is `upsert` or `delete`. Answers 202 with a batch id; a repeated Idempotency-Key returns the first batch with 200.
POST `/sources/{id}/products/batch`	With `operation` `tag`: add, remove or set static tags by name on up to 1,000 products of any source — store, file or API. `create_missing` creates new tags within your plan's limit.
GET `/batches`	List recent batches, newest first. Filter by source, status or date; page with a cursor, up to 100 per page.
GET `/batches/{batch_id}`	One batch: counts, status and the per-item errors.
GET `/sources/{id}/products/{product_id}`	Read one product back, as FeedManager stored it after mapping.
GET `/tags`	Your tags with their product counts; `?kind=static` or `dynamic`.
GET `/tags/{tag_id}`	One tag: name, kind, colour, description and how many products carry it.
GET `/tags/{tag_id}/products`	The products a tag holds — for a storefront's best-seller block or your ERP. Up to 1,000 per page, with the fields you ask for.
GET `/limits`	Your plan's limits and what you have used: products, requests left this minute, queued batches.
GET `/openapi.json`	The API description for Postman, Insomnia and code generators. Public, no key needed.

AI agents

Work with your feeds from Claude or Cursor

FeedManager is an MCP server: Claude, Cursor and other AI agents that speak the Model Context Protocol can read your catalog, feeds and builds, find out why a product is disapproved or missing from a feed, start a sync, a build or a page check, and tag products.

POST https://api.feedmanager.org/mcp — Streamable HTTP, one JSON answer per request, with an API key as the bearer token. It speaks protocol revision 2026-07-28 and the three before it (2025-11-25, 2025-06-18, 2025-03-26).
33 tools, listed below. The key needs the mcp scope plus the scopes of what the agent may do; an agent only sees the tools its key allows. Pro and above; every call counts toward your API requests per minute and is logged.
Agents cannot change sources, feeds, rules or mappings — they tell you what to change in the panel. What they start (a sync, a build, a page check, a tag run) has the same limits as in the panel.
FeedManager's own AI is behind three scopes you switch on yourself: ai:read to see runs and suggestions, ai:write to ask the AI about one product, ai:approve to accept a suggestion. Asking spends your AI credits, up to a daily limit you set per key (500 credits unless you change it); accepting changes what your feeds send. With ai:write alone, the agent proposes and a person decides.
Connectors in the claude.ai and ChatGPT web apps need OAuth, which is not offered yet; Claude Code, Cursor and your own agents work today.

claude mcp add --transport http feedmanager \
  https://api.feedmanager.org/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"
{
  "mcpServers": {
    "feedmanager": {
      "url": "https://api.feedmanager.org/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

MCP tools

What an agent can do

Each row needs the mcp scope plus the scope shown. A tool for a module your plan doesn't include says so instead of returning data.

MCP tools and the API key scope each one needs
Tools	Key scope	What the agent gets
Read `catalog_overview` `search_products` `get_product` `catalog_health`	`products:read`	Product counts, search by text and filters (also by Merchant Center status and issue), every field of one product, and the catalog health score with its field gaps.
Read `list_sources` `get_source`	`sources:read`	Your sources with their schedule, status and product count, and a source's last five syncs.
Read `merchant_center`	`sources:read`	Google's own product statuses today and over 14 days, the most common item issues with an explanation, and account issues with a checklist.
Read `list_feeds` `get_feed` `list_feed_runs` `get_channel` `list_promotions`	`feeds:read`	Feeds and their builds: rejected products with the reasons, what each rule and filter group did, channel and AI readiness, a channel's field specification, and promotions.
Read `explain_product_in_feed`	`products:read` + `feeds:read`	Why one product is or isn't in each feed: the filter group that left it out, the fields the channel rejects, or the row the channel gets.
Read `page_checks_overview` `list_page_checks`	`feeds:read`	Page checks: the daily pass per feed, and per page what the feed sends, what the page shows, the likely cause and what the store answered.
Read `list_alerts`	`sources:read` + `feeds:read`	The alerts that are open now, with their numbers and text; the recently resolved ones on request.
Read `list_tags`	`tags:read`	Static and dynamic tags with their product counts.
Read `list_reports` `get_report`	`reports:read`	Weekly and monthly reports with the numbers behind every section.
Read `list_activity`	`activity:read`	Who changed what, newest first. People appear by name; no email or IP addresses.
Read `list_ai_runs` `get_ai_run` `get_product_ai`	`ai:read`	AI runs and their suggestions; for one product, which AI actions would do something and what each costs in credits.
Read `list_creative_sets`	`creatives:read`	Creative Studio's creative sets with their template, sizes and last render.
Action `sync_source` `build_feed` `check_page` `check_issue_pages` `evaluate_tag` `render_creative_set`	`runs:write`	Start a sync, a build, a check of one product's page or of the pages behind a Merchant Center issue, a dynamic tag's rule, or a creative set's render.
Action `tag_products`	`tags:write`	Add or remove a static tag on up to 1,000 products per call.
Action `improve_product`	`ai:write`	Ask FeedManager's AI about one product: Google category, missing attributes, title, description or an image check. Spends AI credits; the answer is a suggestion.
Action `decide_suggestion`	`ai:approve`	Accept or discard a suggestion. An accepted value is what your feeds send from their next build.

Limits

1,000 products per request; larger bodies are refused before any work is done.
60 requests a minute on Pro, 300 on Premium — one budget per account, shared by all its keys.
Up to 200 queued batches on Pro and 1,000 on Premium; over that you get 429 with Retry-After.
Over the product limit, new products are rejected item by item while existing ones keep updating.

Behaviour worth knowing

The API is incremental: products you don't mention are never touched. There is no "missing means deleted".
A product deleted in the panel is refused per item until you restore it, even if your system keeps sending it.
Request payloads are deleted 7 days after a batch finishes; batch records are kept for 90 days.
Dropping below Pro disables the keys without deleting them; upgrading brings them back.

Push your catalog with the Product API

JSON over HTTPS

1,000 per request

Safe retries

Per-item errors

Getting started

Endpoints

Work with your feeds from Claude or Cursor

What an agent can do

Limits

Behaviour worth knowing

Try it on the Free plan first

Push your catalog with the Product API

JSON over HTTPS

1,000 per request

Safe retries

Per-item errors

Getting started

Endpoints

Work with your feeds from Claude or Cursor

What an agent can do

Limits

Behaviour worth knowing

Try it on the Free plan first

Analytics cookies