Your own model
Run a workspace's AI on your Anthropic key, Bedrock, Foundry, OpenRouter, or a model you host.
Run a workspace's AI on your Anthropic key, Bedrock, Foundry, OpenRouter, or a model you host.
By default Releavo's AI runs on Claude, paid for with the workspace's credits. A workspace admin can point it at a model the workspace pays for itself, on the Team and Pro plans (and on every self-hosted install).
Workspace settings → AI model. Pick one, fill it in, and press Save and test. Releavo sends a short test question that has to be answered with a tool call, because all of its work is tool calls, and shows whether it worked and how long it took.
| What you give it | |
|---|---|
| Your Anthropic key | An API key from the Claude Console. Optionally which Claude models to use. |
| Amazon Bedrock | The region, a Bedrock API key (or an IAM access key and secret), and the model: an inference profile id such as us.anthropic.claude-…. |
| Microsoft Foundry | The resource name, its API key, and the deployment name. |
| Another model | The address of any server with Anthropic's Messages API, its key, and the model. Presets for OpenRouter, DeepSeek, Moonshot (Kimi), Z.ai (GLM) and, self-hosted, Ollama, vLLM, LM Studio and LiteLLM. Press the refresh button next to the model to list what the server offers. |
The model does the careful work (drafting, reviews, code). The quick model answers chat and keeps connected tools in sync; leave it empty to use the same one.
Releavo is built and tested on Claude. Other models can work well, but they follow long instructions and use tools less reliably, so try one on a real task (draft a story, review a doc) before the team relies on it. Pick one that supports tool use and has a large context: 64k tokens or more. With OpenRouter, the model list only shows models that support tools.
Pick Releavo's Claude and save. The stored key is deleted.