Pepedocs

Quickstart

Install Pepe, create an agent, and run the first conversation.

In a few commands you install Pepe, create an agent, and talk to it. pepe setup takes the shortest path: model, key, first agent, and optional channel setup.

1. Install

curl -fsSL https://pepe-agent.com/install.sh | sh
pepe help

2. Configure

pepe setup

The guided setup writes ~/.pepe/config.json. When it asks for a key, prefer a reference like ${OPENROUTER_API_KEY} so the secret stays out of the file.

3. Talk

pepe run assistant "what files are in this directory?"

If you set a default agent, omit the name:

pepe run "summarize the README in three bullets"

For an ongoing conversation:

pepe chat assistant

pepe run answers once and forgets: nothing carries over to the next run. To pick a conversation back up later in the terminal, give the session a name:

pepe chat assistant --session my-session

When a tool wants to act on your machine, such as running a command or writing a file, Pepe asks for your approval first.

4. Serve the API and dashboard

pepe serve --port 4000

The same agent is now reachable in three places:

Test the API:

curl http://localhost:4000/v1/chat/completions \
  -H 'content-type: application/json' \
  -d '{"model":"assistant","messages":[{"role":"user","content":"hi"}]}'
The API starts local. Until you create a token, only this machine can call /v1: nobody else can reach your agent. Create one with pepe token add before exposing the server.

5. Connect a channel

Telegram is the fastest test because it does not require a public URL:

pepe gateway telegram setup
pepe gateway telegram

After that, anyone messaging the bot talks to the same agent. WhatsApp, Slack, Discord, Teams, and Google Chat are covered in Channels.

6. Automate

pepe cron add
pepe watch add "site up" --probe "curl -sf https://example.com" --every 120

Use scheduled tasks for work that repeats, and watches to be told once when something you care about changes.

Next steps

← DockerModels →