I have an AI subscription. Can't I just ask my AI to do what you do?
You can, but then it has to watch the same sources as Orbis itself, manually, instead of using us as a tool for your AI around the clock. Pick the AI you would use and see what it costs.
Which AI would you use?
Claude Sonnet 5.5 is a mid-size model. It handles ordinary agent work and costs less per token than the large models.
Every model gets the same task, the same sources and the same number of calls. Only the price per token differs.
Do it yourself
DKK 172,157/mo
This is the token bill when Claude Sonnet 5.5 has to read everything from the same sources as Orbis itself.
34,852 AI calls and 432.4M tokens per day
With Orbis
DKK 1,568/mo
The Business plan (DKK 1,499) plus DKK 69.43 in tokens for your own AI, which only gets the new hits.
100 hits and 162,000 tokens per day
The AI bill is only part of the job
The figure above only covers your AI reading and judging. Everything else has to be built and kept running too.
- Round-the-clock collectionEvery source fetched on time, at night and at weekends tooDo it yourself: Yours to solveWith Orbis: Included
- DeduplicationThe same story from several sources counts onceDo it yourself: Yours to solveWith Orbis: Included
- Relevance filteringOnly what matches your monitoring gets throughDo it yourself: Yours to solveWith Orbis: Included
- Alerts, history and exportNotified when something happens, and an archive you can search and export fromDo it yourself: Yours to solveWith Orbis: Included
- Servers, development and operationsSomeone has to build it, keep it running and fix it when sources changeDo it yourself: Yours to solveWith Orbis: Included
With Orbis you pay DKK 1,568 a month for the whole monitoring job. That is DKK 170,589 less than the AI bill alone.
Where the spend goes
See every assumptionCalls, tokens per call, prices and the Orbis calculation
Prices for Claude Sonnet 5.5: $2 per million input tokens, $0.2 for cached input and $10 for output.
| Task | Calls per day | New input tokens | Cached tokens | Output tokens | Tokens per day | Per month |
|---|---|---|---|---|---|---|
| Publication checks (211 every 15 minutes) | 20,256 | 7,500 | 5,000 | 1,200 | 277.5M | DKK 113,150 |
| New publications read | 4,500 | 3,500 | 5,000 | 800 | 41.9M | DKK 14,364 |
| Social page checks (4,000 twice a day) | 8,000 | 6,000 | 5,000 | 1,000 | 96M | DKK 36,708 |
| Register checks (every 15 minutes) | 96 | 1,000 | 5,000 | 300 | 605,000 | DKK 115 |
| Register events read | 2,000 | 7,800 | 0 | 400 | 16.4M | DKK 7,820 |
| Total, do it yourself | 34,852 | 432.4M | DKK 172,157 | |||
| With Orbis: your AI gets new hits, batched hourly (100 hits per day) | 24 | 500 | 5,000 | 1,250 | 162,000 | DKK 69.43 |
- Pages are read as cleaned text, which is the cheapest way.
- Tokens per call are estimates based on measured page sizes and include instructions, the previous version of the page and the model's answer.
- Prices are the providers' list prices from October 2026. Where a provider publishes no cached-input price, it is set to 10% of the input price.
- We use DKK 6.65 per USD and 30 days per month.
- Small models may need more calls to reach the same quality. We have not measured that, so it is not included.
- With Orbis your AI is notified of new hits, batched once an hour. An average user gets about 100 hits a day, about 120 tokens per hit. Each time we count 5,000 cached tokens for instructions and 300 output tokens per hit.
Let your AI judge. Let Orbis keep watch.
Orbis watches the web around the clock, removes duplicates and sorts, so your AI only gets the hits that match your monitoring. It spends its tokens judging and acting, not reading thousands of unchanged pages.
Orbis is an infrastructure startup. We build scalable, highly efficient infrastructure that lets AI be lightning fast and keep up with what is happening on the web at all times. We keep expanding into new parts of the internet, giving you more options.