Skip to main content
POST
/
company
/
discover
Discover companies using natural language
curl --request POST \
  --url https://api.datalegion.ai/company/discover \
  --header 'API-Key: <api-key>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "query": "AI companies with 100+ employees founded after 2015",
  "limit": 10
}
'
{
  "matches": [
    {
      "company": {
        "legion_id": "e9f2a4b6-8c3d-4e7f-a1b2-5d6e8f9a0c3b",
        "name": {
          "cleaned": "openai inc",
          "display": "OpenAI",
          "raw": [
            "OpenAI, Inc."
          ]
        },
        "domain": "openai.com",
        "industry": "technology, information and internet",
        "type": "private",
        "size": "1001-5000",
        "founded": 2015,
        "linkedin_url": "https://www.linkedin.com/company/openai",
        "linkedin_id": "17876832",
        "legion_employee_count": 3200,
        "legion_average_tenure": 18.6,
        "legion_employee_growth_rate": {
          "1m": 0.028,
          "3m": 0.085,
          "6m": 0.18,
          "12m": 0.42
        },
        "last_seen": "2026-01",
        "num_sources": 4
      }
    }
  ],
  "total": 562,
  "generated_query": "SELECT * FROM companies WHERE industry ILIKE '%ai%' AND type = 'private' AND founded >= 2010"
}

​
How It Works

Describe what companies you’re looking for in plain English. The API translates your description into a SQL query and returns matching companies. The generated SQL query is returned in the generated_query field of the response body so you can see exactly how your description was interpreted.

​
Example Queries

AI startups in San Francisco with over 50 employees
public companies in the healthcare industry
fast-growing tech companies hiring aggressively
enterprise SaaS companies founded after 2015
companies with high employee turnover in the last year
For precise control over filtering, use Company Search with SQL syntax instead.

​
Pagination

To page through results beyond the first 100, set offset to the number of rows to skip. Maximum offset is 10,000. Ordering is deterministic within a build, so paged requests don’t overlap or skip rows; builds run on a periodic cadence, so paginate within a single client session, not across days. 
{
  "query": "AI startups in San Francisco with over 50 employees",
  "limit": 100,
  "offset": 100
}

Authorizations

​
API-Key
string
header
required

Body

application/json

Request model for natural language company discovery.

​
query
string
required

Natural language description of companies you're looking for

Required string length: 3 - 1000
​
limit
integer
Required range: 1 <= x <= 100
​
offset
integer
default:0

Number of results to skip for pagination (0-10000). Stable within a build; ordering may change across builds.

Required range: 0 <= x <= 10000
​
titlecase
boolean
default:false

If true, format text fields in title case (names, company names, locations). Raw fields, IDs, URLs, codes, and confidence fields are excluded.

​
include_fields
string | null

Comma-separated list of fields to include in response. If omitted, all fields are returned.

​
exclude_fields
string | null

Comma-separated list of fields to exclude from response. Applied after include_fields filter.

​
pretty_print
boolean
default:false

If true, pretty-print JSON response with indentation.

Response

Success - companies found matching natural language query.

Company discover response with generated query.

​
matches
CompanyMatchResponse · object[]
required

List of matches sorted by confidence (descending). Capped at the request's limit.

​
total
integer
required

For /company/search and /company/discover, the total number of rows matching the query's WHERE clause across the database (not just this page). For /company/enrich, the number of matches found for the input identifier. On search/discover, if the count subquery fails or times out, this falls back to the size of the returned page; the response header Total-Count-Status distinguishes the two cases.

​
generated_query
string
required

SQL query generated from the natural language input