Home

🗂 Categories

No results found

OpenCode — AI Coding Agent

If you like this project, consider supporting me on Buy Me a Coffee ☕️

tags:

Description

OpenCode is an open-source AI coding agent — terminal TUI, desktop app, and IDE extension. Multi-provider LLMs, plan/build modes, MCP support, skills, and permission policies. Used in this repo's toolchain. / OpenCode — open-source AI-агент для кодинга: TUI, desktop, IDE-расширение. Мультипровайдерные LLM, plan/build, MCP, skills, политики разрешений.

Common use cases / Типовые сценарии:

Status: Active (anomalyco/opencode). Alternatives: Aider, Claude Code, Codex CLI. / Статус: активно развивается; альтернативы — Aider, Claude Code, Codex CLI.

Paths: config via /connect or opencode.ai/docs/config; project AGENTS.md.
Install: script, npm, Homebrew tap.

Cross-reference: Ollama, LiteLLM, MCP.

Installation

Install OpenCode / Установка OpenCode

# Install script
curl -fsSL https://opencode.ai/install | bash

# npm
npm install -g opencode-ai

# Homebrew (macOS)
brew install anomalyco/tap/opencode
opencode --version
opencode

Configuration

Main files / Основные файлы

AGENTS.md (project instructions)

~/.config/opencode/ (user config)

opencode.json / opencode.jsonc (optional project config)

Provider connection / Подключение провайдера

# In TUI
# /connect  — choose provider (Anthropic, OpenAI, OpenRouter, Ollama, ...)
# /init     — generate AGENTS.md
# /share    — share session

Using local Ollama / С локальным Ollama

# /connect → Ollama → pick model (e.g. llama3.2)
# Or configure provider base URL to http://127.0.0.1:11434

Core Management

TUI commands / Команды TUI

Command Action
/connect Choose LLM provider
/init Create AGENTS.md
/models Switch model
/share Share session
/help Show help
Ctrl+C Cancel / exit

Non-interactive / Неинтерактивный режим

opencode run "explain this repo structure"
opencode -p "list systemd units related to nginx"

Sysadmin Operations

Sessions and logs / Сессии и логи

ls ~/.local/share/opencode/
# Session data under user data dir

MCP servers / MCP-серверы

# Configure MCP servers in project or user config
# See opencode.ai/docs for MCP config format

Permissions / Разрешения

# OpenCode enforces tool permissions (shell, edit, etc.)
# Review permission prompts; use project policies for automation

Security

Hardening / Ужесточение

  1. Do not commit API keys / Не коммитьте API-ключи
  2. Prefer scoped provider keys / Используйте ключи с ограниченными правами
  3. Review shell tool permissions / Проверяйте разрешения shell-инструментов
  4. Use AGENTS.md to constrain agent behavior / Ограничивайте поведение через AGENTS.md
  5. Keep OpenCode updated / Обновляйте OpenCode
  6. Audit MCP server configs / Аудируйте конфиги MCP-серверов
# Keys typically via /connect or environment — never hardcode in repo

Troubleshooting

Symptom Cause Fix
Provider auth failed Wrong/expired key /connect re-auth
Model not found Wrong model id /models select valid model
Ollama not reachable Ollama down / wrong port Check Ollama at :11434
MCP server fails Bad config / binary missing Fix MCP config; install server
Slow responses Large context / small model Trim context; use faster model 
# Check Ollama backend
curl http://127.0.0.1:11434/api/tags
opencode --version

Comparison Tables

AI coding agents / AI-агенты для кодинга

Agent Interface Notes
OpenCode TUI / desktop / IDE Multi-provider, MCP, skills
Aider Terminal Git-native, repo map
Claude Code Terminal Anthropic-focused
Codex CLI Terminal OpenAI-focused

Production Runbooks

Runbook: Onboard OpenCode for a team / Внедрение OpenCode в команде

  1. Install via script or package manager / Установить
  2. Each user /connect to approved provider / Подключить провайдер
  3. Create project AGENTS.md with repo rules / Создать AGENTS.md
  4. Define permission policies for automation / Задать политики разрешений
  5. Document allowed models and key storage / Задокументировать модели и ключи
  6. Optional: route via LiteLLM gateway / Опционально: шлюз LiteLLM

Runbook: Use Ollama as backend / Ollama как бэкенд

  1. Deploy Ollama; pull coding model / Развернуть Ollama; загрузить модель
  2. /connect → Ollama → model / Подключить
  3. Test with small prompt / Проверить
  4. Document model choice and RAM needs / Задокументировать модель и RAM

If you like this project, consider supporting me on Buy Me a Coffee ☕️

Source: devtoolbox-cheats

On this page

linux 🧬 OpenCode Cheatsheet Description Installation Install OpenCode / Установка OpenCode Configuration Main files / Основные файлы Provider connection / Подключение провайдера Using local Ollama / С локальным Ollama Core Management TUI commands / Команды TUI Non-interactive / Неинтерактивный режим Sysadmin Operations Sessions and logs / Сессии и логи MCP servers / MCP-серверы Permissions / Разрешения Security Hardening / Ужесточение Troubleshooting Comparison Tables AI coding agents / AI-агенты для кодинга Production Runbooks Runbook: Onboard OpenCode for a team / Внедрение OpenCode в команде Runbook: Use Ollama as backend / Ollama как бэкенд Documentation Links