---
name: steamwebapi
description: Build integrations with SteamWebAPI for Steam inventories, item catalogs, market prices and CS2 release discovery. Use when implementing or debugging a SteamWebAPI client or choosing its documented endpoints and data fields.
---

# SteamWebAPI integrations

Use the current API contract to implement the user's requested integration. This
skill supplies integration guidance; it does not authorize account changes,
trades, purchases, deployments or other external writes.

## Discover the contract

- With the SteamWebAPI MCP connected, use `search_endpoints`, then `get_endpoint`
  for the selected route. Check required inputs, response schemas and examples
  before generating code. Use `list_games` for game identifiers and `list_markets`
  for current marketplace identifiers.
- Without MCP, read the live OpenAPI document at
  https://www.steamwebapi.com/api/doc.json and documentation at
  https://www.steamwebapi.com/api/doc. Do not invent paths or parameters.
- Treat live endpoint documentation as authoritative when it differs from this
  installed skill. Report unsupported capabilities explicitly.

## Credentials and request behavior

- Keep the SteamWebAPI API key on the backend in environment configuration.
  Do not embed it in browser bundles or print credential-bearing request URLs.
- MCP discovery is available without a key. The optional
  `call_readonly_endpoint` tool uses the connection's configured credentials;
  it only executes allowed documented GET routes. Do not pass Steam cookies,
  Steam Guard secrets or full trade URLs into MCP tools.
- Check HTTP status and parse errors. Handle rate limits with bounded backoff;
  do not retry indefinitely. Match refresh intervals and caching to the product's
  freshness requirements and endpoint quotas.
- Use `select` and documented pagination for the required data. Do not silently
  return only the first page when the user needs a complete catalog or inventory.

## Interpret item and price data

- Use `get_concept` for price and identity fields before calculations. Steam
  listing prices, completed-sale prices and third-party prices are different
  observations. `pricereal` is a third-party market price, not a Steam sale.
- Preserve unknown/null values. Do not turn unavailable prices into zero-value
  items or equate missing metadata with a confirmed negative fact.
- An item catalog entry identifies a market item; it is not a unique inventory
  asset. Use the documented inventory identifiers for asset-specific operations.
- `variants` on items describes phase variants, such as Doppler phases. It is
  not an enumeration of wear conditions or StatTrak/Souvenir versions. Variant
  `pricereal` may be null.

## Discover newly released CS2 items

- `/steam/api/items?game=cs2&only_new_items=1` (also `true`) returns items missing
  from the normal database from all skin collections sharing the newest valid
  release date. It does not return every historical missing item.
- `with_preview_items=1` returns normal items plus those same missing items.
  Preview rows have `preview: true`, a normal-format item ID, collection name in
  `tag7`, and limited metadata. Do not assume all normal price fields exist on
  preview rows. `select` retains the preview marker automatically.
- `releasedat` is the earliest known base-skin collection release date, or null.
  Its object shape is `{"date":"2026-07-08 00:00:00.000000","timezone_type":3,"timezone":"UTC"}`;
  midnight is a presentation convention, not a known release time.
  It does not establish the release date of a specific Souvenir/StatTrak variant
  or its first Steam listing. Do not substitute it for a first-seen timestamp.
- `steamlisting` on preview rows is inferred from a source Steam economy image;
  it is not a live listing-availability check. A preview `inspectlink` can be a
  generated visual certificate with representative wear, not a real owned asset.
- Discovery depends on the upstream catalog and its release dates. An empty
  response is not proof that Valve has released no new content.

## Verify the integration

Use a small representative read to verify fields, types, filters and pagination
before increasing volume. Use `get_pricing` only when quotas or plan selection
matter; size against current endpoint-group limits and the actual workload.
Distinguish verified API output from example data and untested assumptions.