Guides · AI adoption
One central API connected to several AI model providers
Ironhack for Business
- One account per provider
- One key and one invoice
Technical teams now work with several models at once, and every provider wants its own account, its own invoice and its own data policy. OpenRouter AI centralises that access behind a single API. For your company, what matters is which cases it pays off in, and what controls it demands.
What is OpenRouter AI?
OpenRouter AI is a single API that gives access to over 500 models from more than 80 providers with one account, one key and one invoice. Your team still chooses between Claude, GPT, Gemini or open-weight models. What changes is that they are all consumed from the same point.
OpenRouter implements the OpenAI specification for the completions and chat endpoints, so it works as a drop-in replacement for the OpenAI API in applications that already use it.
Why technical teams combine several models
The State of AI report (a16z and OpenRouter, State of AI 2025), published in December 2025 over more than 100 trillion tokens processed on the platform, concludes that no single model or provider dominates usage. Open-weight models reached roughly a third of usage by the end of 2025.
These are OpenRouter's own traffic figures, not the whole market, and the authors present them as observational. They still explain why a technical team working with several models looks for a single access point instead of maintaining one integration per provider.
How OpenRouter works: one key, many models
The flow for a technical team has 4 steps:
OpenRouter · How it works
One OpenAI-compatible API in front of more than 80 providers.
One key for the whole catalogue
Your team creates an account and generates an API key that works for every model.
You pick the model, or routing does
On each request you choose the model, or let automatic routing choose it.
If a provider fails, it tries the next
OpenRouter sends the request to a provider serving that model and, if it fails, uses the fallback.
Usage comes out of your credits
With spend budgets and per-key usage filters.
OpenRouter sits as a single point between your application and more than 80 model providers.
What does OpenRouter AI cost?
OpenRouter AI adds no markup to each model's per-token price: it charges a 5.5% platform fee when you buy credits on the Standard plan. Its FAQ sums this up as «no markup», with a minimum fee of $0.80 per card purchase and 5% on crypto payments.
If your company already has a direct contract with a provider, it can use its own keys inside OpenRouter (BYOK, «bring your own key»). The Standard and Business plans include $25,000 a month of list-price inference with no fee, and charge 5% beyond that.
Your company's invoice therefore has two parts: the per-token price of the model you use, and your plan's fee. These are the differences between plans per OpenRouter's pricing page, consulted in October 2026:
| Plan | Platform fee | Models | EU routing | Data and admin controls | DPA |
|---|---|---|---|---|---|
| Free | N/A | 25+ free | No | No | Not stated |
| Standard | 5.5% | 500+ | No | ZDR and admin controls | Not stated |
| Business | 8% | 500+ | Yes | ZDR and admin controls | Via terms of service, signed copy on request |
| Enterprise | Discounts available | 500+ | Yes | ZDR, admin controls, SSO/SAML and a contractual SLA | Customisable |
Is OpenRouter free?
OpenRouter has a free plan with more than 25 free models, limited to 50 requests a day. If you buy at least 10 credits, the limit on free models rises to 1,000 requests a day. The Free plan is for trying it out, not for company use: it has no data-policy routing, no admin controls and no support beyond the community.
What happens to your company's data on OpenRouter?
By default OpenRouter does not store your prompts or the responses, but it does log metadata for each request (time, model and token count), and each provider it routes to applies its own retention policy. These are the controls worth reviewing before opening a corporate account:
- Prompt logging: you can turn it on in exchange for a 1% discount. With company data, the prudent setting is off.
- Zero Data Retention: ZDR limits routing to providers that do not store the data, and can be enforced on the account, in a guardrail or per request.
- Unclear policies: if OpenRouter cannot establish a provider's policy, it assumes it retains and trains on the data.
- Tools outside ZDR: ZDR only covers inference. Plugins and tools such as web search follow their own retention policies.
- Data in the EU: EU routing is only available on the Business and Enterprise plans. With it, prompts and responses are processed inside the region.
None of these controls replaces a review by your legal and data protection teams. Review the DPA available on your plan (on Business, via the terms of service) and the policy of each provider you intend to use.
When does OpenRouter make sense in a company, and when does it not?
OpenRouter makes sense when a technical team integrates several models into its own products or flows, and it does not replace the Claude, Copilot or ChatGPT licences the rest of the workforce uses day to day. These criteria help decide:
| Criterion | Fits OpenRouter | Better another way |
|---|---|---|
| Who uses it | Developers and product teams | Non-technical employees: AI assistant licences |
| Models | Several models and providers | A single provider on a direct contract |
| Continuity | You need failover between providers | An internal case that tolerates occasional outages |
| Data | You can enforce ZDR and, if needed, buy Business or Enterprise | You require EU residency on the Standard plan |
| Governance | You have a usage policy and a list of approved tools | You do not yet know which AI tools are in use |
OpenRouter · Which plan
It does not repeat the table: it adds the plan choice.
Will technical teams use it?
If not, what you need is AI assistant licences, not an API.
Do you use several models?
If not, a direct contract with one provider works out better.
Do you have a usage policy and approved tools?
If not, the inventory comes first: another API does not tidy up sprawl.
Do you need SSO/SAML or a contractual SLA?
Then Enterprise.
Must the data be processed in the EU?
If yes, Business. If not, Standard with ZDR switched on.
How to decide whether OpenRouter fits, and which plan to choose.
If your engineering teams are going to work with AI models in code, Ironhack for Business's technical workshops run on Claude, Copilot or ChatGPT, depending on your company's stack, and cover generation, refactoring, testing and pair programming.
Access to models is not adoption
OpenRouter solves access to the models, not how your company uses them. If teams already work with several tools and no shared policy, adding an API with over 500 models widens the sprawl and the shadow AI rather than tidying it.
That is why at Ironhack for Business we start with a diagnosis, not a tool: first map what is used, with what data and for which tasks, then decide what access makes sense. Our AI consulting starts with an audit of current use: which tools circulate, unapproved ones included, with what data and at what risk.
It comes in 2 formats, a Lite Assessment (3 weeks) and a Full group assessment (6 to 8 weeks), and includes data and AI governance: usage policy, approved tools and a register. If you want to know what documentary evidence you are missing against the AI literacy obligation, the Article 4 test tells you in 5 questions.
How do you evaluate OpenRouter AI in your company?
Evaluate OpenRouter AI in 4 steps: a tool inventory, a data policy, a bounded pilot and measurement by task.
- Take an inventory: which models and AI tools your teams use today, with what data and on which tasks.
- Define the data policy before creating the account: apply the controls from the data section (account-level ZDR, prompt logging off) and choose the plan by where the data must be processed.
- Run a bounded pilot: 1 technical team, 1 use case and a spend budget set on the account.
- Measure by task: hours before and after, and the real cost with the fee included. To estimate the saving before the pilot, the savings calculator gives an indicative figure.
Is OpenRouter cheaper than contracting the provider directly?
Not per token: OpenRouter applies the provider's inference price with no markup and adds your plan's platform fee. What can pay off is the management: one invoice, failover between providers, and the ability to change model without integrating another API.
Which OpenRouter plan does a company need?
It depends on where the data must be processed and what controls your company requires. Standard works for a technical pilot with ZDR and admin controls, Business adds EU routing, and Enterprise adds SSO/SAML, a contractual SLA and a customisable DPA.
Can you use Claude with OpenRouter?
Yes: Anthropic's models are available on OpenRouter, and the Zero Data Retention settings allow retention rules specific to the Anthropic model group.
Next step
If you want to tidy up AI use in your company before adding another tool, we will come back with a starting point within 1 working day.