Guides
AthenaHQ, Promptwatch or LLM Pulse in 2026: an API key from $95
Promptwatch opens its API on a $95 plan, LLM Pulse on €299, AthenaHQ on a $295 plan plus an unpublished add-on. Rates run $13.61 to $81.94 per 1,000 answers.

By Mathias C.SEO consultant, and more recently GEO, at MentionBox
AthenaHQ, Promptwatch and LLM Pulse are three GEO platforms that also sell an API, and they differ most on where the key sits. Promptwatch prints "MCP and API access" on every paid plan from $95 a month, LLM Pulse limits the REST key to "Scale plans and above" at €299 a month, and AthenaHQ sells API access as a paid add-on on its $295 Starter plan with the price quoted in-app. On the cheapest plan that carries the key, that is $15.83, €30.67 and $81.94 per 1,000 AI answers, checked on October 23, 2026. Our pick for a brand team that wants the data in its own warehouse, after hands-on testing by Mathias C., is API for GEO at $13 per 1,000 answers with no plan to buy at all.
This guide is for a brand team choosing between the three after a dashboard trial, when the figures have to land in a warehouse, a Looker Studio report or a quarterly deck rather than in someone's browser. It compares the three on one table, computes what a realistic prompt list costs on each, maps the metrics field by field across their references, then takes each API in the order of our ranking for enterprise brand teams. MentionBox, which publishes this site, has a commercial link with API for GEO.
The three APIs at a glance
| API | Engines, of the 10 we track | Plan that carries the key | Price per 1,000 answers | Metrics returned | On-demand run |
|---|---|---|---|---|---|
| API for GEO | 10 | none, pay as you go | $13, down to $7 at volume | share of voice, position, sentiment, themes, per-brand perception | yes |
| LLM Pulse | 10 | Scale, €299 a month | €30.67 weekly, €6.65 daily | share of voice, position, sentiment | on prompt creation only |
| AthenaHQ | 10 | Starter $295 plus a paid add-on | $81.94 | share of voice, position, sentiment | yes |
| Promptwatch | 8 | Essential, $95 a month | $15.83, or $13.61 on the $245 plan | position, sentiment, no share of voice | not documented |
Published by each vendor on its pricing page and in its API reference, checked on October 23, 2026. For API for GEO, four of the ten engines, Gemini, Google AI Overviews, DeepSeek and Meta AI, plus the webhooks, the OpenAPI spec and the official SDKs, were confirmed in our hands-on test on October 26, 2026 and are not yet documented on apiforgeo.com.
Four facts decide most of these choices, and none of them is the entry price:
- How the vendor counts. AthenaHQ and API for GEO bill credits, one per AI response. Promptwatch bills a prompt ceiling and a response ceiling at the same time. LLM Pulse bills tracked prompts and lets the refresh frequency set how many answers that buys.
- Who computes share of voice. AthenaHQ and LLM Pulse publish an endpoint for it. Promptwatch publishes none, so the metric a client asks for first is an aggregation you write over its
brandMentionsarray. - Whether you can ask for a fresh answer. AthenaHQ starts a run from the API. The other two hand back what the plan's schedule already collected, which bites on the Thursday a product manager asks about five new prompts.
- What the terms allow. LLM Pulse writes the white-label permission and its limit into its terms. Promptwatch advertises white-label reports without a readable clause, and AthenaHQ's terms are silent on the question.
What one brand team's prompt list costs on each
Take a realistic schedule for a single brand: 100 prompts, 5 engines, 3 markets, refreshed weekly. That is 100 × 5 × 3 × 52 ÷ 12, about 6,500 AI answers a month and 300 prompt slots, since the same question asked in three countries is three tracked prompts almost everywhere.
| Route | What the entry plan with the key covers | Monthly bill for this schedule |
|---|---|---|
| API for GEO | no plan; 6,500 credits sit inside the first tier | $84.50, at $0.013 per credit |
| LLM Pulse | Scale weekly, 450 prompts, 5 models, about 9,750 answers | €299 |
| AthenaHQ | Starter, 3,600 credits, so 2,900 answers short | $533 at the published rate, plus extra credits and the API add-on, both quoted in-app |
| Promptwatch | Essential, 50 active prompts, 6,000 responses | $579 on Business, the first plan with 300 prompt slots |
Our arithmetic on each vendor's published allowances, prices checked on October 23, 2026. Promptwatch's plan ceiling is the binding one: its API returns a 402 reading "Your organization has 50 of 50 active prompts" once the prompt list passes the plan, and a monitor's country is fixed when it is created, so three markets means three monitors.
Two details move that table more than the headline rate. On Promptwatch, cadence is not yours to set: its data freshness page states that "the default, and what every plan uses, is daily", and that the slower options are "enabled per organization by sales, not by picking a plan". Run those 300 prompts daily on 5 models and the formula on its response limits page, prompts times models times runs, gives 45,000 responses a month against the 42,000 the $579 Business plan includes. On LLM Pulse the same lever runs the other way: the €449 daily Scale plan buys the same 450 prompts 30 times a month instead of roughly four, which is 67,500 answers and €6.65 per 1,000, the lowest published rate of the three.
Price per 1,000 answers if you fill the plan
The rate everyone quotes comes from the cheapest plan with a key, which is the worst rate each vendor offers.
| Plan that carries the key | Monthly price | AI answers it covers | Price per 1,000 |
|---|---|---|---|
| API for GEO, pay as you go | none | any volume, credits never expire | $13, then $11, $9 and $7 by tier |
| Promptwatch Essential | $95 | 6,000 responses | $15.83 |
| Promptwatch Professional | $245 | 18,000 responses | $13.61 |
| Promptwatch Business | $579 | 42,000 responses | $13.79 |
| LLM Pulse Scale, weekly | €299 | about 9,750 | €30.67 |
| LLM Pulse Scale, daily | €449 | about 67,500 | €6.65 |
| AthenaHQ Starter | $295 plus the API add-on | 3,600 credits | $81.94 before the add-on |
Prices and allowances published by each vendor, checked on October 23, 2026. The answer counts for LLM Pulse are our arithmetic on its published prompt allowance and its five included models, at 52 weeks or 30 days a month; the others are the vendors' own figures. API for GEO's tiers apply from 50,000, 250,000 and 1,000,000 credits a month.
Promptwatch Professional at $13.61 per 1,000 is the cheapest computed route among the three if your prompt list fits 150 slots, which no comparison we read mentions, because the quoted figure is always the entry plan. The euro figures are not converted here: we have no published rate to convert them with, and a procurement team will want the invoice currency anyway. Our note on the cost of tracking 100 prompts on five engines walks through the same multiplication on a wider field.
Where each metric lives in the payload
The four references name the same measures differently, which is the work nobody does for you before an integration. Our explainer on share of voice, visibility and position covers why two vendors can report the same answers as different percentages.
| Measure | API for GEO | LLM Pulse | AthenaHQ | Promptwatch |
|---|---|---|---|---|
| Answer text | answer_text, responses[] | GET /answers/:id; the list endpoint truncates at 10,000 characters | response, omittable from the payload | content |
| Share of voice | shareOfVoice | GET /metrics/sov | POST /metrics/share-of-voice/cumulative: value, relative_mention_rate, absolute_value, is_self | not published; compute from brandMentions |
| Position | avgPosition | average position on the prompt summary, plus a position distribution report | position, nullable, and competitors[].position | brandMentions[].position, citations[].rank |
| Sentiment | sentiment score, themes, per-brand perception | sentiments inside the full answer, plus a negative-sentiment webhook | attribute observations with a score from -1 to 1 | sentiment, sentimentScore, toneOfVoice |
| Cited sources | sources[] with url, domain, category, mentionsBrand | GET /dimensions/sources, with a source type classification | sources[] with url, root_domain, normalized_url, suffix | citations[] with url, domain, rank, sourceType, contentType |
| Fan-out questions | fan-out endpoints, top, overview and by query | GET /dimensions/query_fan_outs | filterable with has_search_queries, absent from the responses schema | GET /query-fanouts |
| Shopping | shopping products with merchants and price ranges | GET /dimensions/shopping, Scale and above | shopping metrics with top_own_products and merchant_breakdown | shopping product appearances and position analytics |
| Social and video citations | YouTube citations | not published | not published | Reddit, YouTube, LinkedIn and Instagram citations |
Field and endpoint names published in each vendor's API reference, checked on October 23, 2026 and read again on November 16, 2026.
Two of those rows need reading twice. AthenaHQ is the only one of the four that publishes the denominator behind its share of voice: value is the percentage of AI responses mentioning the brand, while relative_mention_rate is "the entry's mentions as a percentage of responses that mention at least one tracked brand". Store whichever you pull, with its name, or next quarter's comparison will mix the two. And on AthenaHQ's fan-out, its own documentation warns that the captured search queries "are searches the model issued while preparing its answer" and that their absence does "not prove the model performed no search", which is the right caution for every vendor in this field. Our explainer on query fan-out covers what those sub-queries are worth.
1. API for GEO, every metric at $13 per 1,000 with no plan
API for GEO is the route here with no subscription at all: one credit buys one prompt, executed once, against one model, in one locale, with the enrichment counted inside that credit. Its pricing page sets $0.013 per credit up to 50,000 a month, then $0.011, $0.009 and $0.007 at the higher tiers, with no upfront commitment, credits that never expire, 24 months of history and reads over stored runs left unmetered, so pulling four quarters into a dashboard costs nothing extra.
Its prompt tracking endpoints follow the shape of a tracking job rather than a dashboard export: POST /v1/queries upserts the prompt list with tags for topic, market and funnel stage, POST /v1/jobs/execute runs it across a model list, and POST /v1/queries/stats reads the result back per prompt.
curl -X POST https://api.apiforgeo.com/v1/jobs/execute \
-H "Authorization: Bearer $API_FOR_GEO_KEY" \
-H "Content-Type: application/json" \
-d '{
"models": ["chatgpt", "perplexity", "google-ai-mode"],
"executionsPerQuery": 3
}'
The job comes back with its id, a status of queued, the models it will hit, the number of queries and the creditsCharged, so the bill is visible before the answers are. The stats call then returns runs, visibility, avgPosition, entities[], sources[], responses[] and lastExecutedAt per prompt.
- Engines and data: all ten we track. ChatGPT, Perplexity, Google AI Mode, Claude, Copilot and Grok are listed on apiforgeo.com; Gemini, Google AI Overviews, DeepSeek and Meta AI were confirmed in our hands-on test on October 26, 2026. One response carries share of voice, position, sentiment, ranked and categorised sources, themes, per-brand perception, fan-out questions, shopping products and YouTube citations.
- Integration: 300 requests a minute and 10 concurrent executions on pay as you go, an MCP server, plus webhooks, an OpenAPI spec and official SDKs confirmed in our test on the same date, with the SDK languages not published. Resale and white label are named as an intended use in its terms.
- Limits, said plainly: $13 per 1,000 is well above the $1 to $4 a scraping API charges for raw answers; the key arrives on request through a form rather than at sign-up; four engines and the whole tooling list come from our test rather than from the vendor's pages, which is a documentation gap rather than a capability we invented; and no SOC 2 report or single sign-on is published, which is where a large procurement team will stop.
- For whom: a brand team that wants the computed layer in its own tables, reads its history as often as it likes and does not want a seat-based platform in the middle. It scores 9.0 of 10 in our ranking of 55 APIs, the highest of our survey, 9.6 in the enterprise ranking, and first in our ranking of GEO metrics APIs.
To try it, request a key on apiforgeo.com: it comes with a sandbox project.
2. LLM Pulse, eighteen engines and the lowest rate at a €299 floor
LLM Pulse has the widest engine list and the strongest integration of the three. All ten engines in our grid and eighteen in total, with Mistral Le Chat, Qwen, MiniMax, Baidu AI, Naver AI Briefing, Brave AI, Alexa for Shopping and Walmart Sparky among them, which settles the choice outright for a brand selling in Korea, China or Japan. Its REST API overview states the gate, "Available on Scale plans and above", the base URL https://api.llmpulse.ai/api/v1, a Bearer key, and a budget of 300 requests a minute per key of which 60 may be writes, shared with MCP calls made on the same key.
curl "https://api.llmpulse.ai/api/v1/metrics/sov?project_id=$PROJECT_ID" \
-H "Authorization: Bearer $LLMPULSE_KEY"
- Price: €299 a month on Scale with weekly tracking, €449 with daily tracking, both for 450 prompts and the five standard models. Our arithmetic on those allowances gives €30.67 and €6.65 per 1,000 answers, so the cadence is the biggest lever on its unit price. Claude, Copilot, Grok, DeepSeek, Meta AI and the eight engines beyond our grid are add-ons from €8 per model per month.
- Returned: the full answer with mentions, citations, sentiments, sources, shopping products, brand entities and fan-out questions on
GET /answers/:id, with the list endpoint truncating content at 10,000 characters. Metrics endpoints cover share of voice, a timeseries by brand and competitor, a prompt summary and the top cited domains; dimensions cover sources, citations, fan-outs, shopping and paid placements, the last two on Scale and above. - Integration: SDKs in twelve languages plus a CLI, an OpenAPI 3.0 spec, HMAC-signed webhooks on Scale and above for new mentions, citations, executions and negative sentiment, bulk creation of up to 100 prompts, pagination capped at 100 rows, and an MCP server over OAuth on every plan including the trial. For a brand team in the European Union, its security annex is the most useful page of the three: "the application, its databases and its job queues run on dedicated servers rented from Hetzner in Germany", with a published sub-processor list and a data processing addendum that "applies to every customer automatically and needs no signature".
- Limits: the €299 floor is the highest entry ticket here, no endpoint re-runs an existing prompt, the collection method is described only in general terms as collecting the answers users see, no country list is published behind an eighteen-engine footprint, and no SOC 2 or ISO attestation and no single sign-on appear on its security page. Full white label, with your own key prefix and a branded API reference, is switched on by its team rather than in the product.
- For whom: a brand team with buyers in markets the rest of this list ignores, a stable prompt list, and a legal department that wants to know which country the servers sit in. It scores 8.8 of 10 overall and 9.5 in the enterprise ranking, second in our GEO metrics ranking.
3. AthenaHQ, on-demand runs and the identity plumbing, at $81.94 per 1,000
AthenaHQ covers all ten engines we track plus Mistral, and it does two things the other two do not. The first is the on-demand run: POST /api/v1/response-streaming/start exists to "run tracked prompts against external AI providers to collect new responses for a website", takes between 1 and 1,000 prompt ids, an optional map of selected models, between 0 and 10 paraphrased prompt_variations, and flags for personas and locations, then returns a workflow_id and a status of running that you poll on the status endpoint.
curl -X POST https://api.athenahq.ai/api/v1/response-streaming/start \
-H "x-api-key: $ATHENAHQ_KEY" \
-H "Content-Type: application/json" \
-d '{
"website_id": "'"$WEBSITE_ID"'",
"prompt_ids": ["'"$PROMPT_ID"'"],
"prompt_variations": 3,
"use_locations": true
}'
The second is identity. Its SSO documentation states that "AthenaHQ supports SAML 2.0 and OIDC and works with any standards-compliant identity provider", that single sign-on comes with Enterprise plans, and that provisioning runs over a SCIM 2.0 endpoint with user create, read, update and deprovision, while noting that group provisioning, bulk operations and sort are not implemented.
- Price: $81.94 per 1,000 answers, our arithmetic on the $295 Starter plan and its 3,600 credits at one credit per AI response, the highest rate of the four. API access on Starter is a paid add-on "billed on top of the Starter subscription" with the amount quoted in the application, so the real cost per answer is higher and unknown. The free Essential plan carries 300 credits and no API.
- Returned:
POST /api/v1/responsesfilters on dates, models, prompt ids, competitor ids, locations, prompt type and schedule, and returnsmentioned,position,cited, theresponsetext,sources[], a compactsource_domainsalternative andcompetitors[]with a position each. Share of voice has its own cumulative and over-time endpoints, sentiment arrives as brand attribute observations scored from -1 to 1 rather than as one field on the answer, and shopping metrics return top own and competitor products with merchant, average position and average rating. - Integration: 60 requests a minute per organization with no cap on the total, an OpenAPI spec, an MCP server with read and write tools, batch prompt creation and dashboard embedding through iframe tokens. No SDK and no webhooks are published, and 60 requests a minute is the tightest ceiling of the four, so a nightly warehouse pull has to be paced.
- Limits: no collection method is published for any engine, no country list is published, and the fan-out questions you can filter on do not appear in the responses payload. Its terms address neither resale nor white label, while its agency page offers discounted licenses to bundle with managed services.
- For whom: a brand team on a modest answer volume that needs ten engines, SAML with SCIM provisioning, and the ability to refresh a prompt the morning a campaign launches. It scores 7.4 of 10 overall, 8.4 in the enterprise ranking and third in our GEO metrics ranking.
4. Promptwatch, the published method and the cheapest key at $95
Promptwatch is the one platform of the three whose API comes with the entry plan, and the one that documents how each answer was collected. Its method page states that "for ChatGPT, Promptwatch runs automated browser sessions that open the actual interface and enter your prompt the same way a user would", that "Gemini, Perplexity, Copilot, Alexa, and Meta AI answers are collected from the live products, and Google AI Overviews and AI Mode from real Google Search results pages", that "API models are called through their provider APIs, on purpose", and that "live-search runs are made from the monitor's country, so a monitor set to Germany collects the answers a user in Germany sees". Mixed methods in one table are a reporting caveat rather than a flaw, and they are written down, which is more than AthenaHQ or LLM Pulse publish per engine.
One GET /responses/{id} on https://server.promptwatch.com/api/v2 returns the computed layer per answer:
curl https://server.promptwatch.com/api/v2/responses/$RESPONSE_ID \
-H "X-API-Key: $PROMPTWATCH_KEY"
The payload carries content, model, provider, visibilityScore, sentimentScore, sentiment, toneOfVoice, mentionedOurBrand, citationCount, a brandMentions array with brandName, brandDomain, a relation of SELF, DIRECT_COMPETITOR, OTHER or IGNORED, plus visibilityScore, position and sentiment per brand, and a citations array with url, domain, rank, title, sourceType and contentType.
- Price and plans: $95 Essential for 50 prompts and 6,000 responses, $245 Professional for 150 and 18,000, $579 Business for 350 and 42,000, Enterprise on quote, with a 7-day free trial and "MCP and API access" on every card. That is $15.83, $13.61 and $13.79 per 1,000 responses respectively.
- Engines and coverage: 8 of the 10 we track as checked on October 23, 2026, Grok and Meta AI absent, on a list that goes past the usual assistants with Amazon Alexa, Mistral, Llama, Codex, OpenCode and Claude Code. Its method page, read on November 16, 2026, now names Meta AI among the live products it collects, so that count may be moving.
- Beyond the answers: the API groups go wider than any other here, with Reddit, YouTube, LinkedIn and Instagram citations, an ads radar, a page tracker, a site audit, content gap analysis and Search Console and Bing Webmaster data, all readable with the same key.
- Limits: no share of voice endpoint, so the portfolio metric is yours to aggregate; no SDK and no webhooks, only a Postman collection and a downloadable OpenAPI 3.0 spec; rate limits set per organization and shared between REST v1, REST v2 and MCP, with an hourly ceiling that starts at 100 requests on the lowest tier; a query window capped at 92 days since October 5, 2026; a monitor's language and country frozen after creation since October 19, 2026; and no security or compliance page we could find, which is a procurement problem rather than a product one.
- For whom: a brand team that wants a key this afternoon for under $100, cares how the answers were collected, and will compute share of voice itself. It scores 7.4 of 10 overall and 7.9 in the enterprise ranking.
What each vendor publishes for procurement
| API | Hosting or residency | Single sign-on | Attestation | Resale or client-facing use |
|---|---|---|---|---|
| API for GEO | not published; sub-processor list published | none published | none published | allowed, named as an intended use |
| LLM Pulse | "dedicated servers rented from Hetzner in Germany" | none published | none published | branded client use on eligible plans; resale as a standalone dataset forbidden |
| AthenaHQ | not published | SAML 2.0, OIDC and SCIM 2.0 on Enterprise | trust centre published, contents not readable without JavaScript | discounted agency licenses; terms silent |
| Promptwatch | not published | none found | none found | white-label dashboard and branded client access advertised; terms page not found |
Published on each vendor's security, documentation, pricing or legal pages, read between October 23 and November 16, 2026. "Not found" marks a page we looked for and could not open, which differs from a vendor publishing nothing.
A team that needs SAML today has one answer in this group, and it costs the Enterprise tier. A team whose legal department asks where the data sits has one answer, and it is in Germany. Both constraints are worth raising in week one, since neither is a feature a vendor can add in the quarter you need it. Our guide on white-label AI visibility APIs covers the clause wording in more detail, and our ranking for enterprise brand teams scores twelve APIs on coverage, data, integration and price.
Five mistakes we see on this shortlist
- Comparing entry prices. $95, $295 and €299 buy 6,000 answers, 3,600 credits and about 9,750 answers. The only comparable figure is the rate per 1,000 at the volume you will actually send.
- Forgetting the prompt ceiling. Promptwatch and LLM Pulse cap active prompts, and a market list multiplies that list. One hundred prompts in three countries is 300 slots, which is two plan steps up on Promptwatch.
- Assuming you control the cadence. Promptwatch runs daily on every plan unless sales enables something slower, and that single setting triples or quadruples the response count the plan has to cover.
- Taking share of voice as one measure. AthenaHQ returns both a percentage of all responses and a percentage of responses mentioning any tracked brand, Promptwatch returns neither and expects you to count, and a figure imported without its definition will not compare with next quarter's.
- Buying coverage you cannot read back. Sixty requests a minute on AthenaHQ, 300 on LLM Pulse shared with MCP, an hourly bucket on Promptwatch shared across every key in the organization. Size the nightly job against the ceiling before the contract, not after.
Our view
Three defensible answers, depending on the constraint that binds hardest.
- You want the computed layer in your own tables, with no platform in the middle. API for GEO is our pick: all ten engines, share of voice, position, sentiment, themes and per-brand perception in one response, $13 per 1,000 answers down to $7, no plan floor, unmetered reads over 24 months of history, resale allowed in writing. The trade-off, in our view, is the documentation gap, since four engines and the tooling come from our test of October 26, 2026, plus the absence of a published attestation or single sign-on.
- You sell in Asia, or your legal team asks where the servers are. LLM Pulse at €299 a month covers eighteen engines from German hosting, with twelve SDKs, signed webhooks and the lowest rate of the three at €6.65 per 1,000 on daily tracking. Budget the add-ons: five models are included and Claude, Copilot and Grok are not.
- You need a key cheaply, or you need SAML. Promptwatch at $95 is the fastest start and the most transparent on method, at the price of computing share of voice yourself. AthenaHQ is the only one here with SAML and SCIM, and the only one that runs a prompt on demand from the API, at $81.94 per 1,000 answers plus an add-on priced in a conversation.
Whichever you shortlist, run your own prompt list on it for two weeks before a figure reaches a deck, and store the raw answers next to the metrics: a change in how mentions are counted then costs a recount rather than a quarter of fresh collection. Our walkthrough of an in-house GEO dashboard built on an API covers the pipeline after the key, and our guide to Profound alternatives with a self-serve API widens the field to the scraping routes at $1.20 to $4 per 1,000 answers.
Frequently asked questions
Which of AthenaHQ, Promptwatch and LLM Pulse includes an API on its entry plan?
Promptwatch. Every paid card on its pricing page prints "MCP and API access", starting at $95 a month for Essential. LLM Pulse states in its REST API overview that the key is "Available on Scale plans and above", which is €299 a month with weekly tracking, while its MCP server works over OAuth on every plan including the trial. AthenaHQ sells API access as a paid add-on on its $295 Starter plan, quoted in-app, and includes it on Enterprise; its free Essential plan has none. Checked on October 23, 2026.
What do the three cost per 1,000 AI answers?
On the cheapest plan that carries the key: $15.83 with Promptwatch Essential at $95 for 6,000 responses, €30.67 with LLM Pulse Scale weekly at €299, and $81.94 with AthenaHQ Starter at $295 for 3,600 credits, before the API add-on whose price is not published. Filling a larger plan lowers the rate: $13.61 on Promptwatch Professional at $245 for 18,000 responses, and €6.65 on LLM Pulse's €449 daily Scale plan, where the same 450 prompts run 30 times a month instead of four. Prices checked on October 23, 2026.
Can you track Gemini brand mentions through these APIs?
All three cover Gemini, on different terms. LLM Pulse includes it among the five standard models on every plan, with ChatGPT, Perplexity, Google AI Mode and Google AI Overviews. Promptwatch states that Gemini answers are collected from the live product and that live-search runs are made from the monitor's country. AthenaHQ lists Gemini among the eleven models on its Starter plan but publishes no collection method, so a Gemini figure from its API carries no stated provenance. None of the three is the Gemini API: grounding with Google Search returns the answer your own call produced, not the one the Gemini app shows a buyer.
Which of the three tracks Claude citations?
All three track Claude, and the method differs. Promptwatch states that a few API models, Claude among them, run with a web search tool attached, and its response object returns a citations array with url, domain, rank, sourceType and contentType for every answer. LLM Pulse sells Claude as an add-on model from €8 per model per month on top of its €299 Scale plan. AthenaHQ includes Claude in the model list of its $295 Starter plan. Checked on October 23, 2026.
Can any of these APIs run a prompt on demand?
AthenaHQ can: POST /api/v1/response-streaming/start runs tracked prompts against external AI providers for between 1 and 1,000 prompt ids, returns a workflow_id, and you poll the status endpoint until it stops reporting running. LLM Pulse queues newly created prompts for execution but documents no endpoint to re-run an existing one, and Promptwatch documents none either, so its monitors deliver on their schedule, which defaults to daily on every plan. API for GEO, our editorial pick after hands-on testing by Mathias C., publishes both: POST /v1/jobs/execute takes a model list and an executionsPerQuery count and returns a job to poll, at $13 per 1,000 answers with no monthly minimum.
Can an agency or a brand team resell the data from these three?
LLM Pulse is the clearest: section 18 of its terms lets you present the service or its data to your clients under your own branding on eligible plans, while section 14 forbids reselling the data as a standalone dataset. Promptwatch advertises a white-label dashboard and branded read-only client access with only the agency invoiced, and we found no terms page at the time of our check, so the clause itself is unverified. AthenaHQ's agency page offers discounted licenses to bundle with managed services, and its terms address neither resale nor white label, deferring to an order form. Read on October 23, 2026.