Company intelligence brief

Give a website and one goal, get back a structured brief on the company that your own AI model or your team can act on straight away: what the company sells and to whom, the market and buyer language, real competitors, what buyers ask and object to in public discussions, and a ready playbook for the chosen goal. Every factual item links to the evidence it came from.

One brief serves one goal:

focus	Built for
reddit_promotion	promoting the company on Reddit: communities that matter, search queries that surface relevant threads, buyer questions and objections, safe claims
seo_web	promoting the company in search: commercial queries by intent, who owns the results today, content gaps
competitor_intel	studying the company as a competitor: positioning, pricing, features, what its customers complain about

A second goal for the same company is a second request. Within 7 days it reuses the site reading and company profile from the first brief, so it starts faster and both briefs describe the company the same way; the job shows "core_reused": true.

Company intelligence is switched on for a key on request: write to support@omev.be. The brief is charged to the key's usual balance.

Depth and price

depth	Price per delivered brief	What is in it
standard (default)	$0.99	the company brief without the audience voice: 3–6 competitors and the standard search; the buyer-voice sections are marked as not included
extended	$1.99	everything in standard, plus the audience voice (up to 300 buyer comments read, with forums and other sources added for narrow niches), 7–8 competitors and a wider search

You pay once per brief, when you first download a finished result, by how complete it is:

Completeness of the brief	Charge
75% or more	the full price
40–74%	70% of the price
15–39%	30% of the price
below 15%	free

A partial charge is rounded down to whole cents. The job shows completeness_percent and charged_usd. A brief that fails, a site we cannot read, a cancelled job and repeated downloads of the same brief are not charged. A domain report includes the extended brief together with community and AI visibility research.

Start a brief

curl -sS https://llmapi.omev.be/v1/cig/jobs/create \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://vessi.com",
    "focus": "reddit_promotion",
    "depth": "standard",
    "client_ref": "acme-client-17",
    "idempotency_key": "acme-17-vessi-reddit-2026-09"
  }'

Field	Required	Meaning
url	yes	the company website; a public http/https address
focus	yes	one of reddit_promotion, seo_web, competitor_intel
depth	no	standard (default) or extended
market	no	a country or region, such as "US", or a list of up to 10, such as ["EU", "DE"]
language	no	a language code such as "en", or a list of up to 6, such as ["en", "de"]
voice	no	who the suggested replies speak as: brand (default) or community_member; OmevAI never posts anything
max_age_days	no	1–1830; discussion threads older than this are left out of the playbook
client_ref	no	your own label for the end client (agencies: split spend per client)
idempotency_key	no	8–200 characters; repeating it returns the same job instead of starting a new one
callback_url	no	a public https address we call when the brief is finished (see Notification)
callback_secret	no	16–200 characters; with it every notification is signed

Before a brief starts, we check that the balance of your key covers its price. If it does not, the answer is 402 INSUFFICIENT_BALANCE and nothing starts.

The answer is 202 with the job:

{
  "job": {"id": "cig_…", "status": "queued", "stage": "", "focus": "reddit_promotion",
          "depth": "standard", "client_ref": "acme-client-17", "charged_usd": 0},
  "idempotent_replay": false,
  "poll_after_seconds": 10
}

Many sites at once

Send up to 50 briefs in one call. Each item takes the same fields as a single brief.

curl -sS https://llmapi.omev.be/v1/cig/jobs/batch \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "items": [
      {"url": "https://vessi.com", "focus": "reddit_promotion", "client_ref": "acme-17"},
      {"url": "https://vessi.com", "focus": "seo_web", "client_ref": "acme-17"},
      {"url": "https://allbirds.com", "focus": "competitor_intel", "client_ref": "acme-17",
       "callback_url": "https://hooks.example.com/omev", "callback_secret": "a-long-random-secret"}
    ]
  }'

The answer is 202 with one entry per item, in the same order. An item with a mistake gets its own error and does not stop the others:

{
  "items": [
    {"index": 0, "job": {"id": "cig_…", "status": "queued", …}, "idempotent_replay": false},
    {"index": 1, "job": {"id": "cig_…", "status": "queued", …}, "idempotent_replay": false},
    {"index": 2, "error": {"code": "INVALID_URL", "message": "…"}}
  ],
  "poll_after_seconds": 10
}

The balance must cover all valid items together; otherwise the call returns 402 INSUFFICIENT_BALANCE and no item starts. Two goals for the same site in one call share one site reading.

Follow the job

A standard brief takes 10–25 minutes, an extended one up to about an hour: the service reads the site, searches, checks competitors and reads buyer discussions. Poll every 15–30 seconds.

curl -sS https://llmapi.omev.be/v1/cig/jobs/get \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"job_id": "cig_…"}'

status is queued, running, then one of completed, partial, failed, cancelled. While running, stage shows where the work is, for example site_read, business_mix, profile_extract, research_plan, market_frame, competitors, search, communities, audience, claim_verify, synthesis, thread_tone, moderation. A brief that reuses an earlier site reading shows site_core_reused instead of the first three. Treat the list as informative: new steps can appear.

partial is a delivered brief with some part not completed; the reason is listed in evidence_base.limitations, and evidence_base.coverage.completeness shows every section with what was found, its target and its minimum. On failed, error.code says why, for example site_unreadable when the site blocks automated visits.

To stop a job, send the same body to /v1/cig/jobs/cancel.

Notification when a brief is ready

With callback_url, you do not need to poll. When the job reaches completed, partial, failed or cancelled, we send a POST to that address:

{"event": "cig.job.finished", "job": {"id": "cig_…", "status": "completed", …}}

Header	Meaning
X-Omev-Event-Id	the job id; use it to ignore repeated deliveries
X-Omev-Signature	sha256= and the hex HMAC-SHA256 of the raw body, keyed with your callback_secret (sent only when you gave a secret)

Answer with any 2xx status. If the call fails, we try twice more, after 10 seconds and after 60 seconds. The outcome is shown in job.callback_status. The notification carries only the job; download the brief with /v1/cig/jobs/result as usual.

Check the signature before you trust the body:

import hashlib, hmac

def is_from_omev(raw_body: bytes, header: str, secret: str) -> bool:
    expected = "sha256=" + hmac.new(secret.encode(), raw_body, hashlib.sha256).hexdigest()
    return hmac.compare_digest(expected, header or "")

Download the brief

curl -sS https://llmapi.omev.be/v1/cig/jobs/result \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"job_id": "cig_…"}'

The response holds job, profile (the full brief) and stage1 (a short summary of the same brief with ten fields). Before the job finishes the call returns 409 JOB_NOT_READY.

The brief has nine sections:

Section	Contents
meta	goal, depth, target URL, market, generation time
company	product, positioning, business mix, features, offer, trust signals
market	buyer profile (ICP), market frame, keywords, buyer queries
problems	problems the brand solves, frustrations in the category
brand_perception	complaints, praise and mentions found in public sources
competition	competitors (direct, adjacent, status_quo) with the reason each one competes
audience	buyer questions, objections, expectations, verbatim buyer quotes
playbook	the plan for the goal: communities, search queries, messaging, next actions, claim guardrails
evidence_base	every source and quote the brief relies on, coverage numbers, limitations

Items carry evidence_ids pointing into evidence_base.evidence, where each entry has the source url and the exact quote. Claims are marked as stated (the company says so), observed (seen in independent sources) or inferred (our conclusion). When you pass the brief to an AI model, tell it to state only stated and observed items as facts and never to use anything listed in playbook.reply_guardrails.do_not_claim.

Errors

Code	When
401	missing or invalid API key
402 INSUFFICIENT_BALANCE	the balance of the key does not cover the brief (or all valid items of a bulk call)
404 JOB_NOT_FOUND	unknown job, or a job that belongs to another account
409 JOB_NOT_READY	the result was requested before the job finished
422 INVALID_URL	the address is not a public website
422 INVALID_FOCUS, 422 INVALID_DEPTH, 422 INVALID_VOICE	a value outside the lists above
422 INVALID_MARKET, 422 INVALID_LANGUAGE, 422 INVALID_MAX_AGE, 422 INVALID_IDEMPOTENCY_KEY	a field out of range: more than 10 markets, more than 6 language codes, an age outside 1–1830 days, a key outside 8–200 characters
422 INVALID_CALLBACK_URL, 422 INVALID_CALLBACK_SECRET	the notification address is not a public https address, or the secret is not 16–200 characters
422 INVALID_REQUEST, 422 BATCH_TOO_LARGE	a bulk call without items, with an item that is not an object, or with more than 50 items
429	too many requests; retry after a short pause

Every error has the same shape: {"error": {"code": "INVALID_FOCUS", "message": "…"}}. A 401 for a wrong key and a 429 work the same way for every OmevAI API; see Errors for their format.