Getwello
All integrations
Local LLM · maximum privacy

Mum's data, your machine, no cloud.

Run a local LLM with Ollama, point an MCP-capable frontend at our open-source MCP server, and you can ask your loved one's check-in data conversationally without a byte leaving your machine. Privacy-maximalist Coordinators only.

Setup guideGet an API key

Coordinator-only. Included in the £4.99 plan. Same day to set up.

~/ollama run llama3.1
$ ollama run llama3.1
Loaded model · MCP server connected · Getwello tools available
> how is Mum doing this week?
Calling Getwello MCP server...
Last 7 days: 6 of 7 check-ins on time. Mood mostly fine,
one low day on Thursday. James visited Tuesday, Chloe Friday.

All data stayed local.
Why a privacy-first family would do this

Local model, local data, local everything.

01

Data never leaves the house

Model runs on your hardware. Prompts and responses stay on the machine.

02

No AI subscription

Open-source models, free to use. Pay only the electricity bill.

03

MCP standard

Same Getwello MCP server as Claude Desktop. We don't care which AI calls it.

04

Good enough for the questions you'll ask

'Has Mum checked in?' is well within a 7B model's capability when given the tool to look it up.

$ ollama run llama3.1
>>> How is Mum doing this week?

[Calling get_circle_stats...]
[Calling list_gap_days...]

Adherence 86% over 7 days (6 of 7 on time).
Mood mostly fine, one low day on Thursday.
Two upcoming visits, gap on Wednesday.
The Thursday dip lines up with the GP
appointment, worth checking in about that.

All data stayed local.
The terminal answer

ollama run, ask, done.

Through an MCP-capable Ollama frontend, the model uses our MCP server like any other. The terminal-style transcript shows tool calls being made, data flowing in, the model interpreting it. Most importantly: every line of that conversation stays on your machine.

  • Run Ollama with any model that supports tool-calling.
  • Use an MCP-capable frontend (Open WebUI, LM Studio).
  • Same npx invocation as Claude Desktop sets up the server.
  • No cloud round-trips except the Getwello API calls themselves.
How to set it up

The high level. Full walkthrough in the help guide.

  1. 01
    Step 1 of 4

    Install Ollama and pull a tool-calling model.

    Download Ollama from ollama.com. Pull a model: ollama pull llama3.1. Confirm it runs: ollama run llama3.1.

  2. 02
    Step 2 of 4

    Install an MCP-capable frontend.

    Open WebUI is the most polished option. Run via Docker: docker run -d -p 3000:8080 ghcr.io/open-webui/open-webui:main.

  3. 03
    Step 3 of 4

    Add the Getwello MCP server to the frontend.

    Most frontends accept the same npx invocation as Claude Desktop. Add the snippet, paste your API key, save.

  4. 04
    Step 4 of 4

    Ask a question.

    Type 'How is Mum doing this week?' The model calls our tools, summarises the result, hands it back. All without leaving your hardware.

Read the Ollama setup guide →Mint your API key

Frequently asked questions

Does Ollama support MCP?

Ollama itself is a model runner. MCP support comes through frontends that speak MCP: Open WebUI, LM Studio, Continue and a handful of others. The Getwello MCP server stays the same, you point an MCP-capable frontend at both.

Which model is best?

For tool-calling, a model trained for function calling is essential. Llama 3.1 8B and above, Qwen 2.5 7B, or Mistral Nemo all work well. Smaller models tend to hallucinate tool calls.

Do I need a beefy machine?

8GB+ of unified memory on an M1/M2/M3 Mac runs the 7-8B parameter models comfortably. On Windows or Linux, a recent GPU helps but is not strictly required for the smaller models.

Does Getwello see what I ask the local model?

No. The local model handles the conversation, then calls our MCP tools (which call our REST API) only for the specific data the model decides to fetch. We see only the tool calls themselves, not your prompts.

Why bother if Claude is so much better?

Privacy. The model never sees your data outside your machine, our API responses never enter any cloud-AI's context. Worth the quality trade-off for some families.

Other integrations

Pair this with another.

Some families wire two or three at the same time. The webhook fans out, the API keys stay the same.

See every integration or jump to the Developer API overview.

Connect Getwello to Ollama.
Today.

Mint your first API key in two minutes. Coordinator-only. Included in the £4.99 family plan.

Start your free month →
£0 todayFirst month on usCancel any time, in one tap
Download on the App StoreGet it on Google Play