# AI use cases for Sprawling Agents (2youg1/sprawling-agents)

Repo: [2youg1/sprawling-agents](https://github.com/2youg1/sprawling-agents) | Use cases: 4 | Page: https://dedupedai.com/use-cases/2youg1-sprawling-agents/

sprawling is a local Rust multi-agent harness where agents live as residents of a city, a directory on your machine, and coordinate through messages, plans and an append-only Ledger.

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

- Run an agent team on a long task while you step away: https://dedupedai.com/use-cases/2youg1-sprawling-agents/run-agent-team-on-long-task/
- Turn a project idea into a roadmap that buildings execute: https://dedupedai.com/use-cases/2youg1-sprawling-agents/plan-project-with-hierarchical-roadmap/
- Review how agents interacted during a run: https://dedupedai.com/use-cases/2youg1-sprawling-agents/replay-agent-interactions-for-review/
- Measure run time, model cost and memory of a city: https://dedupedai.com/use-cases/2youg1-sprawling-agents/measure-city-cost-and-resource-use/

## How to run an agent team on a long task while you step away with Sprawling Agents

Repo: [2youg1/sprawling-agents](https://github.com/2youg1/sprawling-agents) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README

Start a local city, give the Mayor a goal, and let residents work until it is done.

**Who it is for:** A developer with a multi-step coding or research task who wants agents to keep working after they stop watching the screen.

**The problem:** Single agent sessions stall at quota limits and need constant attention, so long tasks wait on you.

**The outcome:** A running city where the Mayor plans the work, residents carry it out, and you check results when you return.

### Steps

1. Install the package with `npm install --global sprawling@latest`.
2. Start the city with `sprawling up ./cities/first` and let the page open in your browser.
3. On the page, connect a provider or a local model and choose a model for `main`.
4. Tell the Mayor the result you want and how far counts as done.
5. Follow progress and answer residents' questions; type `/quit` to close the city when no runs remain, or stop them when asked.

### Prompt to paste into your agent

```text
In the 2youg1/sprawling-agents repo, install sprawling with npm and run `sprawling up ./cities/first` in a terminal. Report the serving address and pairing code the terminal prints, and confirm the address loads in a browser before you stop.
```

### Caveat

sprawling is labelled alpha and its README says data formats, the wire and the interface may still change between versions.

Risk, credentials: Provider keys you paste go into the city vault, but you should still check which account and limits each key carries before leaving runs unattended.

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/2youg1-sprawling-agents/run-agent-team-on-long-task/

## How to turn a project idea into a roadmap that buildings execute with Sprawling Agents

Repo: [2youg1/sprawling-agents](https://github.com/2youg1/sprawling-agents) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README

Have the Mayor write a roadmap and hand each building its part through plan.

**Who it is for:** A small team lead who has a project idea but no written plan, and wants the work split across buildings with a standing goal.

**The problem:** Ideas stay in chat and nobody tracks which part is ready to start or who owns it.

**The outcome:** A hierarchical plan in the city, with ready nodes dispatched and in-progress runs visible to you.

### Steps

1. Start the city with `sprawling up ./cities/first`.
2. Press Esc in the terminal to open the city's CLI and talk to the Mayor.
3. Ask the Mayor to write the idea down as a roadmap and hand each building its part through `plan`.
4. Set a standing goal so the plan nodes that are ready get dispatched.
5. Type `/help` to list the slash commands for steering, pausing and checking the work.

### Prompt to paste into your agent

```text
In the 2youg1/sprawling-agents repo, start a city with `sprawling up ./cities/first` and ask the Mayor to draft a roadmap for the project described in <project brief>. Expected output: a hierarchical plan with one part per building. Verify that each plan node names an owning building before you accept it.
```

### Caveat

The README says the city has no built-in workflows, so you define the workflow yourself through roles, skills and tool connections.

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/2youg1-sprawling-agents/plan-project-with-hierarchical-roadmap/

## How to review how agents interacted during a run with Sprawling Agents

Repo: [2youg1/sprawling-agents](https://github.com/2youg1/sprawling-agents) | Difficulty: Advanced | Time: 1 hour | Basis: From the README

Export the city's message history with playback and check it against the Ledger.

**Who it is for:** A researcher or developer studying how a group of agents coordinates without human relay.

**The problem:** Agent-to-agent messages are hard to audit after the fact, so claims about who did what are hard to verify.

**The outcome:** An exported history of agent messages you can check against the Ledger record.

### Steps

1. Start the city with `sprawling up ./cities/first`.
2. Let residents message one another, find each other and coordinate a task.
3. Ask the Mayor to run the playback skill described in crates/city/skills/playback/SKILL.md to export the history. (inferred, check the README)
4. Open the exported history and compare a sample of messages with the Ledger entries.

### Prompt to paste into your agent

```text
In the 2youg1/sprawling-agents repo, read crates/city/skills/playback/SKILL.md, then export the message history of the city at <city path>. Expected output: an exported history file. Verify that three messages picked at random appear in the Ledger with matching content.
```

### Caveat

The README describes playback as exporting history for checking against the Ledger, so the exact export format should be read from the skill file, not assumed.

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/2youg1-sprawling-agents/replay-agent-interactions-for-review/

## How to measure run time, model cost and memory of a city with Sprawling Agents

Repo: [2youg1/sprawling-agents](https://github.com/2youg1/sprawling-agents) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Read timings, tokens, cost and resource use with gauge and set a memory ceiling.

**Who it is for:** A developer on a shared or limited machine who needs to know what an agent team costs in tokens and memory.

**The problem:** Long runs can use model budget and memory without showing where it went.

**The outcome:** A report of run timings, model calls, tokens, cost and resource use on your hardware, with a memory limit you chose.

### Steps

1. Start the city with `sprawling up ./cities/first` and run a short task.
2. Run `sprawling gauge` to read timings, model calls, tokens, cost and resource use.
3. Open Settings → Performance and set a memory ceiling for each run if you want one.
4. Compare the readings with the budgets in tools/xtask/budgets.toml.

### Prompt to paste into your agent

```text
In the 2youg1/sprawling-agents repo, run a short task in a city at ./cities/first, then run `sprawling gauge` and save its output. Expected output: a summary of run timings, tokens and cost. Verify that the figures are from your own machine, not from the recorded readings in tools/xtask/budgets.toml.
```

### Caveat

The README says there is no memory ceiling by default and only you can set one, so runs have no limit until you choose one.

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/2youg1-sprawling-agents/measure-city-cost-and-resource-use/