autoplay-sdk

Transform your existing customer support chatbot into a proactive, personalized experience for every user.

Documentation · Quickstart · Discord · PyPI

---

🤔 The problem

The problem with reactive customer support chatbots:

• They wait to be asked — assuming users will speak up when they're stuck. They rarely do.
• When users do ask, they're expected to know how to frame their question correctly. But users don't know what they don't know about your platform, and often initiate the conversation in the wrong way.

✅ The solution

What we propose you build with our SDK:

• Don’t wait for users to come to your chatbot. Go to them first — with the right help, at the right moment, personalized to what they’ve been doing in your platform.

• Don’t just tell them how to fix it. Show them — by triggering contextual visual guidance through smart tooltips or a browser agent (coming soon).

---

⚙️ How it works

You have all the tools in your tech stack — you just need to orchestrate them correctly:

1. session replay provider
2. chatbot provider
3. visual guidance provider

---

✨ Features

Feature	What it does
Real-time event stream	Captures and normalizes browser activity into clean, structured data your model can actually use
Context management	Automatically buffers and summarizes high-volume events so you never blow your context window
Golden paths	Record your product's ideal user journeys once — your agent always guides users the right way
Per-user memory	Tracks what each user knows, where they got stuck, and what they've done — so your agent never repeats itself
Proactive triggers	Detects the right moment to intervene and fires a chat message or visual nudge automatically
Interruption control	A built-in state machine ensures your copilot only speaks when it should — helpful, never annoying

---

🚀 Quick Start

Requirements: Python 3.10+

pip install autoplay-sdk
# or
uv add autoplay-sdk

The fastest way to get set up is with the AI agent skills. Run this once from your project root after installing:

autoplay-install-skills
# or target your specific stack:
autoplay-install-skills --chatbot intercom --user-activity posthog
autoplay-install-skills --chatbot ada --user-activity fullstory

This drops a .cursor/skills/ folder into your project. Then open Cursor or Claude and say "Set up Autoplay with Intercom and PostHog" — the agent handles the full wiring automatically: session scoping, conversation linking, and context assembly for your stack.

→ Full setup guide on developers.autoplay.ai/quickstart

The quickstart covers the PostHog frontend snippet, product registration, and streaming your first live events in under 10 minutes.

---

🤖 Chatbot tutorials

Step-by-step guides for wiring Autoplay into your existing chatbot platform.

	Platform	Tutorial
	Intercom	View tutorial →
	Ada	View tutorial →
	Botpress	View tutorial →
	Dify	View tutorial →
	Tidio	View tutorial →
	Landbot	View tutorial →
	Crisp AI	View tutorial →
	Rasa	Coming soon
	Inkeep	Coming soon

---

💬 Community

Discord	discord.gg/jCbR2tQA5 — share what you're building, get help, stay updated
GitHub Issues	Best for bug reports and feature requests
Support	See SUPPORT.md for channels and response expectations
Docs	developers.autoplay.ai
PyPI	pypi.org/project/autoplay-sdk

---

SDK Structure (0.7.4+)

As of 0.7.4, the SDK is organized into a domain-based layout. If you're upgrading from an earlier version, your existing imports continue to work — you'll see a DeprecationWarning guiding you to the new paths.

New import structure

# Core primitives
from autoplay_sdk.core.models import ActionsPayload, SlimAction
from autoplay_sdk.core.metrics import SdkMetricsHook

# Chat
from autoplay_sdk.chat.chatbot import BaseChatbotWriter
from autoplay_sdk.chat.chat_pipeline import compose_chat_pipeline

# Context & memory
from autoplay_sdk.context.agent_context import AsyncAgentContextWriter
from autoplay_sdk.context.context_store import AsyncContextStore
from autoplay_sdk.context.user_index import UserSessionIndex
from autoplay_sdk.context.summarizer import AsyncSessionSummarizer

# Proactive triggers
from autoplay_sdk.proactive.triggers import ProactiveTriggerRegistry
from autoplay_sdk.proactive.triggers.defaults import DEFAULT_COOLDOWN_S
from autoplay_sdk.proactive.resilience.resilience import ProactiveResilienceStore
from autoplay_sdk.proactive.state.idle_expiry import run_proactive_idle_expiry

# Agent state
from autoplay_sdk.agent_state.v1.states import AgentStateMachine  # five-state FSM
from autoplay_sdk.agent_state.v2.states import SessionState        # three-state FSM

# RAG
from autoplay_sdk.rag.ingest.ingest import RagPipeline
from autoplay_sdk.rag.query import assemble_rag_chat_context

# API / FastAPI
from autoplay_sdk.api.fastapi import build_fastapi_app

# Admin
from autoplay_sdk.admin.onboarding import onboard_product

Top-level imports are unchanged and never require migration:

from autoplay_sdk import AgentStateMachine, RagPipeline, ProactiveTriggerRegistry

Upgrading from 0.7.3 and earlier

Old import paths still work in 0.7.4 and emit a DeprecationWarning with the exact replacement. They will be removed in 1.0.0.

To migrate automatically, install the migration skill and run it in Cursor or Claude:

autoplay-install-skills --migrate

Then open your project and say:

"Migrate my autoplay-sdk imports to 0.7.4"

The agent will find every deprecated import, rewrite it to the canonical path, and verify no deprecated paths remain.

Full migration reference → docs/sdk/migration-0.7.4.mdx

📋 Contributing

• Contributor guide: CONTRIBUTING.md
• Code of conduct: CODE_OF_CONDUCT.md
• Security reporting: SECURITY.md

---

📄 License

MIT — see LICENSE.