# Competitor research

> Anonymous Tier C competitor-research workflow for one company domain and 1 to 5 competitor domains.

Source page: https://docs.replynodes.com/docs/guides/competitor-research

# Competitor research

The gateway exposes a read-only competitor-research workflow on the primary API host:

- `GET https://api.replynodes.com/v1/competitor-research?domain=<domain>&competitors=<host,host,...>`

OpenAPI operation ID: `anonymous_competitor_research`.

`domain` is the company, for example `stripe.com`. `competitors` is a comma-separated string of 1 to 5 competitor domains, for example `netlify.com,cloudflare.com`. Both are bare public hostnames: no scheme, port, path, or credentials. A missing, invalid, or blocked domain, or an empty, invalid, or oversized competitor list, is rejected before admission and does not consume the daily bucket.

A valid request admits exactly once, then performs a server-owned, read-only fan-out over url-to-markdown, brand-kit, and brand-logo primitives for the company domain plus each competitor domain. The whole composition is charged once. Those primitives are never charged a second anonymous bucket. The response is a composed JSON dossier with one entry per target.

```bash
curl -sS "https://api.replynodes.com/v1/competitor-research?domain=stripe.com&competitors=netlify.com,cloudflare.com"
```

## Anonymous quota (Tier C)

Keyless callers get a daily allowance per trusted client IP for the `competitor-research` capability. This bucket is separate from `company-research`. The gateway returns `X-RateLimit-Limit`, `X-RateLimit-Remaining`, and `X-RateLimit-Reset` (Unix seconds for the next UTC midnight). A `429` also returns `Retry-After`. Redis or quota degradation returns `503`.

Presenting an API key or OAuth access token moves the call onto the authenticated path; package credit pricing for this workflow is not set yet (see capabilities discovery).

## Response

`200` returns `{ data, meta }` and includes the quota headers. `data.workflow` is `competitor-research`, `data.domain` is the company domain, and `data.targets` has one entry per target. Each target includes `domain` and, when that primitive succeeds, `brand`, `styleguide`, `logo`, and `markdown.content`. A failed primitive is omitted and recorded under `errors` rather than invented. `502` with code `upstream_unavailable` means every primitive read failed after the request was admitted. `400` means the input was rejected before admission.

## MCP

The MCP tool name matches the OpenAPI operation ID:

- `anonymous_competitor_research` → `GET /v1/competitor-research`

MCP tool calls require a ReplyNodes API key or OAuth access token issued for MCP
(`https://mcp.replynodes.com/mcp`). The tool forwards to the same gateway route above.

## Skill attribution

Downstream producers may send `X-ReplyNodes-Skill: company-research` on workflow or
primitive calls so `api_request` analytics carry `skill=company-research` and
`surface=skill`. Only that exact header value is accepted; unknown values are ignored.

## Related guides

- [Company research](https://docs.replynodes.com/docs/guides/company-research) for the single-domain workflow
- [Brand intelligence](https://docs.replynodes.com/docs/guides/brand-intelligence) for the primitives composed inside the workflow
- [Web context](https://docs.replynodes.com/docs/guides/web) for crawl/scrape primitives agents may chain manually
