> For the complete documentation index, see [llms.txt](https://mads.avelino.run/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://mads.avelino.run/how-to/ollama-local.md).

# Run with Ollama

This page shows how to run mads against a local Ollama server, so no data leaves your machine and no API key is needed.

## Steps

1. Install and start Ollama. It listens on `http://localhost:11434` by default.
2. Pull a model that supports tool calling.
3. Run mads with `--provider ollama` and the model name.

```bash
ollama pull <model-name>
mads generate business.toml --provider ollama --model <model-name>
```

`mads providers` shows `ollama` as `ready` always, because it needs no key. It does not test the connection.

## When the server is not running

mads reports the connection error as a failed plan mission and exits with code `1`. This is real output with no server on the port.

```
12:28:31 [run] 20261001-122831-70203f started (ollama, llama3.1)
12:28:31 [plan] started (attempt 1)
12:28:34 [plan] failed: Web call failed for model 'ollama::llama3.1 (adapter: Ollama)'.
Cause: Reqwest error: error sending request for url (http://localhost:11434/api/chat)
12:28:34 [out] out3/20261001-122831-70203f/report.md
12:28:34 [run] failed (exit 1): 0 in, 0 out tokens, cost n/a, 1 of 1 missions failed
```

Start the server and run `mads generate --resume <run-dir>` with the same provider flags. See [Resume and export](/guides/resume-and-export.md).

## A server on another machine

`--provider ollama` always uses `http://localhost:11434`. Ollama also serves the OpenAI chat completions API under `/v1`. Point `openai-compat` at it.

```bash
mads generate business.toml \
  --provider openai-compat \
  --base-url http://<host>:11434/v1 \
  --model <model-name>
```

See [Use an OpenAI-compatible server](/how-to/openai-compatible.md).

## Model quality

mads needs a model that calls tools reliably and keeps long instructions in view. Small local models can break the ad text limits (30 characters per headline) or loop on validation errors. That shows up as `max turns` or `no progress` failures. Raise `--max-turns`, use a larger model or give the plan mission a stronger model with `--plan-model`.

Local runs report tokens but no cost.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://mads.avelino.run/how-to/ollama-local.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.