Skip to main content
Open source - Apache-2.0

Talk to your agents.
The localhost AI Agent runtime.

Chat UI, tools, RAG, and MCP in one pip install. Multi-provider models, voice, images, and skills -- running on your own machine by default.

Get startedTry the live demoView on GitHub
Microsoft Agent FrameworkReact + FastAPIAzure OpenAI - Anthropic - OpenAI - Foundry
ChatWalaʻau analyzing an image in chat
Multi-providerMCPRAGVoiceImagesSkillsOpenAI-compatible APICLI
Capabilities

Everything an agent needs, in one install

Chat, tools, skills, MCP, and RAG -- composed into a single runtime you control.

Localhost-first runtime

A full AI agent server in one pip install. Your chats, files, and keys stay on your machine -- no mandatory cloud.

Bring your own models

Azure OpenAI and Anthropic (Claude), with mid-conversation multi-model switching and per-message reasoning effort.

Rich, readable chat

Streaming replies with Markdown, syntax-highlighted code, KaTeX math, Mermaid diagrams, collapsible reasoning, and slash commands with completion.

Tools & Skills

Coding, weather, and image tools as native functions, plus portable file-based Skills loaded on demand -- with approval gating.

MCP & MCP Apps

Connect external tools over the Model Context Protocol, and render interactive UIs inside sandboxed iframes.

RAG over your docs

Ingest PDFs into a local ChromaDB vector store and let the agent answer with grounded, cited sources.

Voice & multimodal

Speak with Whisper transcription, hear replies with text-to-speech, and analyze images via drag-and-drop or URL -- or draw one in the built-in paint canvas.

Generate & edit images

Create and edit images inline with a Canvas mask editor -- powered by Azure OpenAI image models.

Identity & memory

A global agent identity and a self-maintaining user-preference memory that learns and reconciles your preferences over time -- a built-in learning loop.

Scheduled execution

Schedule scripts on a cron expression, an interval, or once after a delay -- managed from a portal, the API, or the agent itself, with a run timeline and logs.

See it in action

A chat that renders everything

The same rendering quality that powers these docs powers the product.

Multimodal image analysis
Multimodal image analysis
Mermaid diagrams & math
Mermaid diagrams & math
Image generation & editing
Image generation & editing
Rich tool result cards
Rich tool result cards
Searchable session history
Searchable session history
Agent DevUI for debugging
Agent DevUI for debugging
Models

Bring your own models

Run on Azure OpenAI, Anthropic (Claude), OpenAI, and Microsoft Foundry, switch models mid-conversation, and dial reasoning effort per message -- all behind one selector.

Configure providers
  • Azure OpenAIgpt-5.5, gpt-5.4, gpt-4o, o3, gpt-4.1-mini ...
  • AnthropicClaude Opus / Sonnet / Haiku - Direct or Foundry
  • OpenAIgpt-5.x / o-series - direct public API, API key
  • Microsoft FoundryReasoning model deployments - project endpoint, Entra ID
Quickstart

Running in under a minute

1

Install

pip install chatwalaau
2

Configure

chatwalaau init, then set your provider in .env
3

Run

chatwalaau -> open http://localhost:8000/chat

Build it for your business

Self-host the open-source runtime, or let us tailor it to you -- custom deployments, legacy migration, and ontology-driven agents, delivered end to end.

Get startedWork with us