Skip to content

3. Add a model key

Maitento calls models from providers such as OpenAI, Anthropic and DeepSeek on your account with that provider. Add your provider’s API key once, and every run uses it.

The tutorials use DeepSeek’s deepseek-v4-flash, which is fast and inexpensive. Get a key from the DeepSeek platform and add it. For the knowledge and memory tutorials, add an OpenAI key as well: they need embeddings, which DeepSeek doesn’t offer.

Open Settings → Model keys, choose Add model key, pick the provider, give it a label and paste the key. Only tenant admins can add keys.

The Add a model key dialog: provider, label, API key and priority

The page lists your keys, with only their last four characters, and explains the order a run tries them in:

The Model keys page, listing keys by provider with their label, priority and status

await mt.modelKeys.create({ provider: "DeepSeek", label: "Tutorials", apiKey: process.env.DEEPSEEK_API_KEY! });

The providers are OpenAi, Anthropic, DeepSeek, Gemini, Qwen, Glm and Kimi.

  • A key can’t be read back once it’s saved: Maitento only ever shows its last four characters.
  • Several keys for one provider are tried in priority order, so a refused or exhausted key passes the call to the next one.
  • A run with no key for its model’s provider fails straight away and says so. Its failure details end with [api:model-keys:none-for-provider], and the portal shows a button to add one.

More in Models and model keys.

Next: 4. Install an SDK or the CLI.