# AI use cases for Maki (tontinton/maki)

Repo: [tontinton/maki](https://github.com/tontinton/maki) | Use cases: 3 | Page: https://dedupedai.com/use-cases/tontinton-maki/

maki is a terminal AI coding agent built to use as few context tokens as possible while staying fast and pleasant to use. It parses code with tree-sitter to read file skeletons, runs tools from a sandboxed interpreter, and lets subagents pick a weak, medium or strong model. It is written in Rust with a 60 FPS TUI, Lua plugins like neovim, and supports many providers including Ollama and llama.cpp.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Install Maki and cut the tokens your coding agent spends: https://dedupedai.com/use-cases/tontinton-maki/install-maki-and-code-with-fewer-tokens/
- Run a coding agent on your own Ollama or llama.cpp server: https://dedupedai.com/use-cases/tontinton-maki/run-a-local-coding-agent-on-ollama-or-llama-cpp/
- Replace a headless Claude Code script with Maki: https://dedupedai.com/use-cases/tontinton-maki/swap-claude-code-for-maki-in-a-headless-script/

## How to install Maki and cut the tokens your coding agent spends with Maki

Repo: [tontinton/maki](https://github.com/tontinton/maki) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Install the Maki coding agent and use its index and code execution tools to read less and do more per call.

**Who it is for:** A developer whose agent sessions burn through tokens by reading whole files.

**The problem:** Agents that read entire files and call tools one by one fill the context window fast and cost more.

**The outcome:** A coding agent that reads file skeletons first and batches tool calls, with full visibility into what it reads.

### Steps

1. Download the installer with `curl -fsSL https://maki.sh/install.sh -o install.sh` and read it with `cat install.sh`
2. Run it with `chmod +x install.sh && sh install.sh`, or use `brew install maki`
3. Set your provider key, for example OPENAI_API_KEY or GEMINI_API_KEY
4. Start Maki in your project and give it a task; watch the index tool read skeletons before full files

### Prompt to paste into your agent

```text
Install tontinton/maki, configure it with <provider>, and run it on <repo> to <task>. Report how many lines it read and which tools it used.
```

### Caveat

This card is written from the tontinton/maki README, not from a test run by Deduped. Before you start, make sure you have Linux, macOS or Windows. Check the repo for newer instructions before you rely on these steps.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/tontinton-maki/install-maki-and-code-with-fewer-tokens/

## How to run a coding agent on your own Ollama or llama.cpp server with Maki

Repo: [tontinton/maki](https://github.com/tontinton/maki) | Difficulty: Intermediate | Time: 20 minutes | Basis: From the README

Point Maki at a local model so code never leaves your machine and there is no API bill.

**Who it is for:** A developer with a local GPU who wants an agent that works offline.

**The problem:** Many coding agents assume a cloud provider and make local models awkward to wire in.

**The outcome:** Maki talking to a local Ollama or llama.cpp endpoint through a plain environment variable.

### Steps

1. Install Maki with the script or Homebrew
2. For Ollama, set OLLAMA_HOST, for example http://localhost:11434
3. For llama.cpp, set LLAMA_CPP_HOST, for example http://localhost:8080
4. Start Maki, pick the local model, and use the task tool's weak, medium or strong choice for subagents

### Prompt to paste into your agent

```text
Set up tontinton/maki to use my local <Ollama or llama.cpp> server at <host>, pick the best coding model I have, and run a small refactor in <repo> to test it.
```

### Caveat

This card is written from the tontinton/maki README, not from a test run by Deduped. Before you start, make sure you have Linux, macOS or Windows. Check the repo for newer instructions before you rely on these steps.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/tontinton-maki/run-a-local-coding-agent-on-ollama-or-llama-cpp/

## How to replace a headless Claude Code script with Maki with Maki

Repo: [tontinton/maki](https://github.com/tontinton/maki) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run Maki without the UI and keep the same stream-json output your scripts already parse.

**Who it is for:** A team with CI or automation built on Claude Code's print mode that wants to try another agent.

**The problem:** Switching agents usually means rewriting the parsing code around them.

**The outcome:** Your existing pipeline running Maki with compatible JSON output.

### Steps

1. Install Maki on the machine that runs your script
2. Replace the agent call with Maki using `--print --output-format stream-json`
3. Run the script on a test task and compare the output with the old run
4. Optionally turn on the OpenTelemetry export, which uses the same format as Claude Code's

### Prompt to paste into your agent

```text
Change my <script> that runs Claude Code in print mode to call tontinton/maki with --print --output-format stream-json instead, run it on <test task>, and diff the outputs.
```

### Caveat

This card is written from the tontinton/maki README, not from a test run by Deduped. Before you start, make sure you have Linux, macOS or Windows. Check the repo for newer instructions before you rely on these steps.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/tontinton-maki/swap-claude-code-for-maki-in-a-headless-script/