chatpharo

ChatPharo Documentation

ChatPharo is an AI coding assistant integrated into the Pharo IDE. This documentation set explains how it works internally, the features you can use day‑to‑day, and the memory/tooling/safety subsystems.

Start here

1. Open ChatPharo
   • World menu: ChatPharo → Temp ChatPharo
   • Or open ChatPharo Settings / Setup Wizard from the same menu.
2. Configure an agent (local or cloud) in Settings → Agent.
3. Send a message from the chat input.
4. (Optional) Enable Tools, Skills, Memory, Multivers, Sandbox, or MCP in settings.

Documentation map

• Big picture: architecture.md
• End-to-end runtime path: message-flow.md
• Practical usage: features.md + tutorial.md
• Extending ChatPharo: tools.md, skills.md, mcp.md, director.md

Conventions used here

• Class/Package names appear in monospace.
• Diagrams are rendered as plain-text blocks so everything stays Markdown-only.
• Examples use Smalltalk (Pharo).

---

Quick links

• Architecture
• Message flow
• Features
• Configuration
• Agents
• Tools
• Skills
• Memory
• Multivers
• MCP
• Sandbox
• Safety & Ethics
• Director/Team (multi-agent)
• IDE integration
• Tutorial
• Troubleshooting
• Glossary

---

Documentation files

File	What it covers
index.md	Entry point, navigation, and the full file index table.
architecture.md	Layered architecture diagrams and subsystem overview.
message-flow.md	Send message → prompt building → model call → tool loop → UI update.
features.md	User-facing feature tour: chat, browser integration, attachments, feedback, logs, help.
configuration.md	Settings model, defaults, persistence, and how toggles affect behavior.
agents.md	Backends (Claude, Gemini, Ollama, LM Studio, DeepSeek, etc.), capabilities, configuration.
tools.md	Built-in tools, custom tools, tool calling protocol, and execution lifecycle.
skills.md	Skills registry, manual/auto skill context injection, adding skills.
memory.md	Long-term memory model, retention, auto-summarize, prompt context injection.
multivers.md	Multi-step chain execution and prompt rewriting across agents.
mcp.md	MCP registry, connecting servers, and exposing MCP tools to the LLM.
sandbox.md	Code-execution sandbox, restrictions, and timeout behavior.
safety-ethics.md	Safety advisor warnings + ethics budgets/gating for high-impact settings.
director.md	Director/Team pattern for multi-agent task decomposition and result merging.
integration.md	Pharo IDE integration: world menu, editor actions, system browser context.
tutorial.md	Setup and the interactive tutorial entry points.
troubleshooting.md	Common issues and diagnostics.
glossary.md	Terminology used across the docs.

Generated on 2026-02-13.

This site is open source. Improve this page.