Next:   [Contents] [中文] [Source]

Documentation Overview

Welcome to the Magent documentation. Magent is an Emacs Lisp AI coding agent with multi-agent architecture, permission-based tool access, and LLM integration via gptel.

Documentation Index

Getting Started

  • COMMANDS.org — Slash commands, internal LLM workflows, and their management commands
  • ONBOARDING.org — New developer onboarding guide with architecture overview, guided tour, and complexity hotspots
  • TROUBLESHOOTING.org — Common issues and solutions

Architecture

  • ARCHITECTURE.org — Product positioning, system boundaries, module layers, request flow, and capability model
  • AGENT_WORKFLOW.org — Thread/turn/item state machine, loop flow, persistence (snapshot + journal), UI projection, and Codex alignment
  • AGENT_JOBS.org — Durable child-agent job lifecycle, tool surface, persistence, UI, and boundaries
  • UI_BACKENDS.org — Supported agent-shell + ACP flow and frontend development boundary
  • DOCTOR.org — Safe Doctor probe API, trust boundary, redaction, sessions, and cancellation
  • PTC_MULTI_MODEL_PLAN.zh.org — Chinese implementation plan for PTC, run_code, and multi-model phases in one agent loop
  • PROCESS_DISPLAY_PLAN.zh.org — Chinese implementation plan for Codex-style progress paragraphs and pre-tool assistant text in chat continuations

Contribution

  • CONTRIBUTING.org — Contribution guidelines, code style, testing, and PR process
  • RELEASING.org — Versioning policy, public compatibility boundary, changelog rules, and release checklist
  • TROUBLESHOOTING.org — CI, live smoke, and melpazoid failure notes

Project Root Documentation

  • ../README.org — Main project README with features, installation, and usage
  • ../CHANGELOG.org — Release history and current unreleased changes
  • ../AGENTS.md — Development guide for agentic coding tools with build commands, architecture notes, and repository conventions

Quick Links

For New Contributors

  1. Start with ONBOARDING.org
  2. Read CONTRIBUTING.org for development workflow
  3. Read ../AGENTS.md for current architecture notes and development guidance
  4. Read AGENT_JOBS.org before changing child-agent lifecycle behavior

For Users

  1. ../README.org — Installation and configuration
  2. COMMANDS.org — Built-in slash commands and Magent-owned LLM workflows
  3. TROUBLESHOOTING.org — Common issues and solutions
  4. Run M-x magent-action-run-doctor for self-diagnostics
  5. Use M-x magent-start to open the supported agent-shell UI

For Developers

  1. CONTRIBUTING.org — Code style and PR process
  2. ../AGENTS.md — Build commands, testing, architecture notes, and development guidance
  3. ARCHITECTURE.org — Current architecture and system boundaries
  4. ONBOARDING.org — Guided code tour and complexity hotspots
  5. AGENT_JOBS.org — Current child-agent job architecture
  6. UI_BACKENDS.org — Current frontend support boundary
  7. DOCTOR.org — Doctor data boundary and extension contract
  8. RELEASING.org — Versioning and release process

CI And Packaging

  1. ../README.org — Public workflow badges and development commands
  2. CONTRIBUTING.org — Local and CI verification sequence
  3. TROUBLESHOOTING.org — Known GitHub Actions failure signatures
  4. ../AGENTS.md — Agent-facing test, coverage, live smoke, and melpazoid notes
  5. RELEASING.org — Release gates and artifact checks

Documentation Standards

When adding new documentation:

  • Place user-facing docs in project root (README.org)
  • Place developer docs in docs/
  • Use Org for docs under docs/; generated HTML is build output under _site/.
  • Add paired English/Chinese pages and reciprocal magent_alt_url metadata unless a page is intentionally language-specific.
  • Update this index and magent-docs--navigation when adding public pages. The docs builder rejects duplicate URLs, missing navigation entries, language mismatches, and non-reciprocal explicit translation links.
  • Keep stable docs or active task notes updated before stopping work so another machine can resume from git