OpenRouter

OpenRouter puts models from many labs behind one account, and Memtro connects to it by signing in rather than by a pasted key. Connect once and every model OpenRouter serves you can answer wherever Memtro calls a model: chat, agent mode, routing ladders, memory extraction and jobs.

Connect your account

You need an OpenRouter account. Requests are billed to that account's own credits; OpenRouter's FAQ describes credits as deposits for inference, and says free models are limited to a small number of requests a day, raised once you have bought credits.

On Dashboard → Providers pick OpenRouter. There is no key field. Choose who the connection is for: Personal (only you), Shared with your organisation, or one team. Which of these you are offered depends on your permissions. Work done for the organisation or a team, such as its jobs and its connections' syncs, spends only shared and team keys, never a member's personal one; your own personal jobs and personal connections can use your personal key.

Press Sign in to OpenRouter →, sign in there and approve. OpenRouter issues an ordinary OpenRouter API key for your account, Memtro stores it encrypted, asks which models it can reach, and brings you back to the page you started on with OpenRouter connected — <n> models available. Finish within ten minutes of pressing the button, or start again. If the sign-in does not complete you see one of these:

Message	What to do
That OpenRouter sign-in did not complete. Try connecting again.	Start again from the Providers page.
OpenRouter refused to issue a key (…). Try connecting again.	OpenRouter's own reason follows the status; start again.
Could not reach OpenRouter to finish connecting: …	A network failure between Memtro and OpenRouter; try again shortly.
Create or join an organisation before adding a model key.	Every key in Memtro belongs to an organisation.

Memtro asks for no spend limit, label or expiry on the key it receives. To stop using it, press Remove beside it on the Providers page; that removes it from Memtro, and the key itself remains on your OpenRouter account until you delete it at openrouter.ai/keys.

Model ids

Id	What it is
openrouter/<vendor>/<model>	One model through OpenRouter, keeping OpenRouter's vendor prefix: e.g. openrouter/anthropic/claude-sonnet-5, openrouter/openai/gpt-5.4 or openrouter/deepseek/deepseek-chat. Suffixes OpenRouter uses, such as :free, pass through.
auto/openrouter	Routed by difficulty within OpenRouter, on a default ladder that crosses vendors: DeepSeek V3 (deepseek/deepseek-chat) → Gemini 3.8 flash → gpt-5.4 → Claude Sonnet 5, all through OpenRouter. See model routing.

The openrouter/ prefix is required: without it, anthropic/claude-sonnet-5 goes to an Anthropic key rather than to OpenRouter. The model list beside the connection shows everything your account can reach, less the models that do not chat (embedding, speech, moderation, rerank and image models). A named model is used only where an administrator has opened model choice on the Routing page; otherwise it is answered as memtro (see cost control).

What is different about OpenRouter in Memtro

• One shape for every model. Memtro talks to OpenRouter in the OpenAI chat format whichever lab's model answers, so the extras it gives a provider directly do not apply: Claude through OpenRouter gets no Anthropic prompt caching and no native tool search, and no model gets a thinking or effort setting. A reasoning_effort on the request, and agent mode's request for deeper thinking, have no effect here.
• Memtro identifies itself. Every request names Memtro and www.memtro.com to OpenRouter, which reserves some models, its free tier among them, for apps it recognises.
• Free models and memory. Some free models ignore a requested output schema. When Memtro's own structured work, such as memory extraction and the routing classifier, lands on one, it asks again in plain text and checks the answer itself, so a free model on a rung does not stop memory learning.
• No embeddings. An OpenRouter key does not embed memories. Memory uses an OpenAI or Google key your organisation holds, or Memtro's own embedding model.
• More than one connection. Your own is tried first, then your team's, then the organisation's; within each, the order set on the Providers page, or Spread the load. A rate-limited or exhausted key hands over to the next before routing moves down the ladder. See several keys for one provider.
• Usage in tokens. Each request is recorded with the input and output tokens OpenRouter reports. Allowances are set in tokens, and Memtro does not price them: your OpenRouter account is where the spend is.

Your data at OpenRouter

As of November 2026, OpenRouter's FAQ says it does not log prompts or completions by default, only metadata, unless you opt in. It tries to route only to providers that do not log or train on your data, and does not route to providers that do, or cannot be confirmed not to, unless you turn on the model training setting in your OpenRouter privacy settings. What the lab behind a model keeps is set by that lab, so read its terms too. Bear in mind that a request through Memtro usually carries your organisation's memory with it. See privacy for what erasure in Memtro cannot reach.

What this is not

There is no way to paste an OpenRouter key or set a base URL on the Providers page: the sign-in is the only way in. The connection is billed to your OpenRouter credits, not to a subscription with any of the labs behind it.