Before you begin

Install Ollama using its official instructions and download a model that fits your computer. Keep the model identifier returned by your local server. Memory requirements, speed and image support vary by model, so this guide does not prescribe a model that may be unsuitable for your hardware.

Meetingly uses a Chat Completions-compatible endpoint for answers. Ollama documents this compatibility in its OpenAI compatibility guide. Provider billing and usage limits apply when the endpoint is a hosted service; a Meetingly subscription is not required for the app’s own-key answer path.

1. Confirm that the local server is ready

Start Ollama, download your chosen local model, and confirm that it appears in the model list. On the usual local configuration, the API base address is:

http://localhost:11434/v1

You can inspect the model list without sending meeting content:

curl http://localhost:11434/v1/models

Use the exact model ID in that response. If you use a different port or a server on another machine, replace the base address with the address you intentionally configured. Keep a local server bound to your own machine unless you need network access and have secured it.

2. Connect it in Meetingly

  1. Open the Meetingly menu from the tray/menu-bar icon.
  2. Choose Use your own API key….
  3. Choose the Ollama preset or enter the compatible base URL.
  4. Leave the key empty for a standard local Ollama server that does not require a key.
  5. Load the models and choose the exact model you downloaded.
  6. Save. Meetingly makes a short validation request before accepting the configuration.

The endpoint needs both the model-list route and Chat Completions for the complete setup flow. Point at the API base URL, rather than a web dashboard or an individual chat page. The address above ends in /v1.

3. Check the rest of the content path

Choose local transcription in Meetingly and let its speech model finish downloading. If you do not want meeting history uploaded, leave the account disconnected. Local AI answers, local transcription and account sync are independent choices.

Send a harmless test such as “Reply with OK” from the Answers panel. Then use a short sample session to verify that the transcript appears and the model can answer a question about it. Expect local generation speed to depend on the computer and context length.

Common setup problems

Connection refused or cannot reach the host
Check that Ollama is running and that the port in Meetingly matches it. A local endpoint on another computer will not be reachable through this computer’s localhost address.
The model list is empty
Download a model first, then reload the list. Check that the endpoint supports /v1/models.
The selected model returns 404
Copy its ID from the model list. A display name and an API model ID may differ.
Text works but screenshots fail
Choose a model that supports image inputs. A text-only model cannot inspect a screenshot.
Replies are too slow
Try a smaller model suited to your available memory, close competing workloads, and reduce unnecessary prompt context. Test before the meeting.

Switching back to hosted answers

Clear the own-key configuration to return to Meetingly’s hosted answer route, then connect an account with available plan usage. Clearing the endpoint does not delete your Ollama models. For a full overview of data flows, read local vs cloud transcription.