Skip to content

AI-Parrot

Async-first Python framework for building AI Agents and Chatbots.

AI-Parrot is a vendor-agnostic framework that lets you build conversational agents, tool-using assistants and multi-agent crews on top of any major LLM provider — OpenAI, Anthropic, Google GenAI, Groq, VertexAI, HuggingFace — through a single async interface.

Get started → Browse the API →

Documentation by chapter

  • Foundations

    Core abstractions, data models and the architectural decisions that keep AI-Parrot async and vendor-agnostic.

    → Open chapter

  • LLM Clients

    One AbstractClient interface for every provider. Streaming, retries, presets and embeddings.

    → Open chapter

  • Bots & Agents

    Chatbot, Agent, AgentCrew — single agents and multi-agent orchestration with sequential, parallel and DAG execution.

    → Open chapter

  • Memory & Knowledge

    Conversation memory, episodic memory and RAG over PgVector, FAISS, Milvus, Arango or BigQuery.

    → Open chapter

  • Tools, Loaders & RAG

    The @tool decorator, toolkits, OpenAPI ingestion and document loaders for the RAG pipeline.

    → Open chapter

  • Integrations & Transport

    Telegram, MS Teams, WhatsApp, voice. MCP servers/clients and the A2A inter-agent protocol.

    → Open chapter

Quick navigation by use case

  1. Install AI-Parrot
  2. Read the Bots & Agents overview
  3. Pick the tools you need in Tools, Loaders & RAG
  4. Tune behaviour via Configuration
  1. Read Bots & Agents and pick the right execution mode (sequential / parallel / DAG).
  2. For cross-host or cross-process agents, jump to A2A Communication.
  3. Production patterns live in Advanced Orchestration.
  1. Pick a vector store — Storage Backends.
  2. Wire loaders → Local Knowledge Base, Loaders Metadata.
  3. Tune retrieval with Parent-Child Retrieval.
  1. Decide the surface — REST? MCP? Messaging?
  2. REST: API Endpoints.
  3. MCP: MCP Sessions and Simple MCP Server.
  4. Telegram / Teams / WhatsApp: see the Integrations chapter.

Contributing

Documentation lives in docs/ and follows the Style Guide. The site is built with MkDocs Material and the API reference is generated by mkdocstrings from the docstrings in packages/ai-parrot/src/parrot/.

Run the site locally:

pip install -r requirements-docs.txt
pip install -e packages/ai-parrot
mkdocs serve

External resources