> ## Documentation Index
> Fetch the complete documentation index at: https://stampede.autonomousorganizations.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Stampede

> Stampede runs AI assistants on your own machine or server, each with its own permissions, reachable from a terminal, Telegram, webhooks or a schedule.

Stampede runs AI assistants on a computer you control: your laptop, or a Linux server. You give each assistant a name, a job and a set of permissions, then ask it for work from wherever you are. Your files, keys and history stay on that machine.

## What Stampede is

An assistant in Stampede is called a person. A person has a name such as `research` or `builder`, a short description of its purpose, and a policy that says which tools it may use and which programs it may run. One person can read files and nothing else; another can edit code and run your tests. Each person keeps its own memory, goals and mail between conversations.

You reach your persons in four ways:

* From a terminal, in a console where you type requests and watch the work happen.
* From Telegram, where members of a small group write to a bot and a person answers in the chat.
* From signed webhooks, so an event in GitHub, GitLab or Linear (a labelled issue, a new merge request) starts a run.
* On a schedule, so a person does a task every morning or every hour.

Persons can hand work to each other. A person can split a job into smaller pieces and run them in parallel, or send mail to another person and get the answer back.

Every piece of work is written to a journal on disk before each step. If the machine crashes or the process is killed, the record of what finished is still there, and you can resume the run.

Stampede talks to Anthropic, OpenAI, and any server that offers an OpenAI-compatible API, such as Ollama or vLLM running on your own hardware. It runs on macOS and Linux.

## What you can do

* Ask a quick question from the terminal and get one answer: `stampede -prompt "..."`.
* Open the console in a project directory and let the model read files, edit code and run commands.
* Create specialist persons with narrow permissions, and address one by starting a line with `@name`.
* Give a small team a Telegram bot that routes each message to the right person.
* Turn issues and merge requests into runs, with the result posted back as a comment.
* Run a task on a timer and read the result later.
* Run long tasks headless with `stampede agent "..."` and resume them after an interruption.
* Pair two machines so a person on one can ask a person on the other for work.
* Check every part of an installation with `stampede doctor`.

## Where to start

> [card] [Getting started](https://stampede.autonomousorganizations.ai/quickstart.md)
>
> Install Stampede, get a first answer, and create a first person.

> [card] [Concepts](https://stampede.autonomousorganizations.ai/concepts.md)
>
> Home, person, preset, run, sub-agent and gateway, with one example each.

> [card] [Persons](https://stampede.autonomousorganizations.ai/persons/overview.md)
>
> Create persons, give them a purpose and a policy, and address them by name.

> [card] [Telegram](https://stampede.autonomousorganizations.ai/start/telegram.md)
>
> Let a small group reach your persons through a Telegram bot.

> [card] [Webhooks](https://stampede.autonomousorganizations.ai/start/webhooks.md)
>
> Start runs from signed GitHub, GitLab and Linear events.

> [card] [Server install](https://stampede.autonomousorganizations.ai/production/server-install.md)
>
> Run Stampede on a Linux host from a release binary.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.