FOR AGENT BUILDERS · MCP + OPENAPI + SDKs

Competitor intelligence for AI agents

Whatever framework your agent runs on, SimilarIQ plugs in: a hosted MCP server for MCP-native agents, an OpenAPI spec for tool-calling frameworks, zero-dependency SDKs for custom code, and a clean JSON REST API underneath it all.

MCP-native agents (Claude, Cursor, custom MCP clients)

{
  "mcpServers": {
    "similariq": {
      "type": "http",
      "url": "https://similariq.ai/mcp",
      "headers": { "X-API-Key": "sk-live-your-api-key" }
    }
  }
}

Tool-calling frameworks (LangChain, OpenAI tools, CrewAI...)

Point your framework's OpenAPI toolkit at the spec:

https://similariq.ai/openapi.json

Or define one function that wraps a single POST:

POST https://similariq.ai/api/v1/search
Headers: X-API-Key: sk-live-...
Body: {"domain": "salesforce.com", "limit": 100}
→ {"competitors": [{"rank": 1, "company": "HubSpot", "domain": "hubspot.com", "matchScore": 97, ...}]}

Custom code

npm install similariq-sdk   # Node 18+, TypeScript types
pip install similariq       # Python 3.8+, stdlib only

Agent design notes

The tools it gets

10 tools: find_similar_companies, start_bulk_search, get_bulk_job, list_recent_searches, get_search_results, list_my_companies, save_company, list_saved_companies, create_share_link, get_search_balance. Only searches cost credits (1 per domain); everything else is free. Full details on the MCP page.

Frequently asked questions

What does SimilarIQ return for a domain?

The top 100 competitors, look-alikes, and similar companies for any company website — each with a 0-100 match score, relationship type (Direct, Indirect, Adjacent, Emerging), a confidence score, and a likely-competitor flag.

How much does it cost?

1 credit per search (up to 100 results). Plans from $6/month ($0.10-$0.24 per search) or pay-as-you-go at $0.50 per search. Free trial searches on signup; failed searches auto-refund.

Is there an OpenAPI spec?

Yes — https://similariq.ai/openapi.json describes the v1 endpoints with schemas, ready for LangChain, the OpenAI tools format, and any OpenAPI toolkit.

How should agents handle the 30-60 second search time?

Give search tools a 120-second timeout, or use the async bulk endpoints (start a job, poll status) for anything more than a couple of domains.