Skip to main content
Write the Pipecat coding-agent guide into a project so an AI coding assistant (Claude Code, Codex, …) follows Pipecat conventions automatically — then has it scaffold the app for you with pipecat create. pipecat init writes three files:
  • AGENTS.md — the agent-facing guide (how to build Pipecat apps: scaffold first, check APIs against live sources instead of stale training data, verify with the eval harness). Picked up automatically by coding agents.
  • CLAUDE.md — a one-line import of AGENTS.md so Claude Code loads the guide.
  • GETTING_STARTED.md — developer-facing guidance: how to set up the Pipecat Context Hub, write an effective first prompt, and what to expect from a coding session.
pipecat init does not scaffold a project — it makes a project agent-ready. To scaffold a runnable bot, use pipecat create (the recommended flow is to run pipecat init, then let your coding agent drive pipecat create).
Usage:
pipecat init [TARGET_DIR] [OPTIONS]
Arguments:
TARGET_DIR
string
Directory to make agent-ready. pipecat init my-bot targets ./my-bot; pipecat init . targets the current directory. With no argument, the CLI prompts for a directory.
Options:
--force
boolean
default:"false"
Overwrite an existing CLAUDE.md. By default CLAUDE.md is written only when absent, so a developer’s own file is never clobbered.

Behavior

  • AGENTS.md and GETTING_STARTED.md are always (re)written, so re-running pipecat init refreshes the guide after a Pipecat upgrade.
  • CLAUDE.md is preserved if it already exists — pass --force to overwrite it.
  • The written AGENTS.md ends with a provenance footer naming the pipecat-ai version that wrote it, so a stale guide is detectable and refreshable.
  • Passing scaffolder flags (e.g. --name, --bot-type) prints a message redirecting you to pipecat create and exits non-zero.

Examples

Prompt for a directory

pipecat init

Make a specific directory agent-ready

pipecat init my-bot

Make the current directory agent-ready

pipecat init .

Refresh the guide, overwriting CLAUDE.md

pipecat init my-bot --force

Next Steps

Open the project in your coding tool and let the agent take it from here — it reads AGENTS.md and scaffolds the app with pipecat create.

pipecat create

Scaffold a runnable bot — services, pipeline, and deploy path wired up