OpenCode: Detailed Introduction

[ikaijua]

OpenCode: Detailed Introduction

An open-source AI agent with terminal and desktop interfaces and a choice of model providers.

• Audience: People who want model choice while working on local projects and files.
• Cost: Install the client yourself; model or hosted-service usage is a separate cost.

Official account & key contributors on X

• OpenCode (@opencode)

Table of Contents

Getting started

1. What it is and who it is for
2. Access and setup: your first session

Usage and configuration

3. A concrete task and how to check it
4. Costs: software, models, and execution
5. Configuration and extensions

Limits and references

6. Permissions, data, and limitations
7. Frequently asked questions
8. Official references

Use the contents to jump to each section. If an in-page link does not respond, open it in a new tab.

[ikaijua]

What it is and who it is for

In one sentence: An open-source AI agent with terminal and desktop interfaces and a choice of model providers.

Who it suits: People who want model choice while working on local projects and files.

What you can do:

• Explore a repository before changing it.
• Read and edit files for development tasks.
• Extend workflows with custom agents and tools.

Where to start: Choose the terminal client or desktop download from OpenCode.

Cost: Install the client yourself; model or hosted-service usage is a separate cost.

[ikaijua]

Access and setup: your first session

Choose the terminal client or desktop download from OpenCode.

The current repository documents this terminal package:

npm i -g @opencode/cli@latest

Connect a model using the official setup guide, then start in a test project.

[ikaijua]

A concrete task and how to check it

Start with plan to identify entry points and tests, then switch to build for a small change.

Example prompt:

Inspect the login flow without editing. Identify three relevant files and existing tests. Wait for my confirmation, then implement the agreed change and report the checks actually run.

Review the diff, test output, and unfinished work separately.

[ikaijua]

Costs: software, models, and execution

Separate client access, your chosen model/API, and optional hosted services. Check which account is billed when changing providers; a chat subscription does not automatically cover every API.

Measure one small task before committing to long sessions or parallel work. Official docs

[ikaijua]

Configuration and extensions

The built-in build and plan agents are switched with Tab. plan restricts edits by default and asks before Bash; build is the execution surface.

@general handles complex searches and multi-step work. MCP connects external tools: add only the tools needed for the task and inspect their access. Agent docs

[ikaijua]

Permissions, data, and limitations

Plan restrictions are not an operating-system sandbox. Recheck file and command access when switching to build or enabling external tools.

A local client can still send project context to the selected model provider. Open source does not mean entirely local processing.

Tool-calling reliability and long-session behavior vary by model.

[ikaijua]

Frequently asked questions

Is the desktop app required? No; choose the terminal or desktop interface.

Does plan edit files? It restricts edits by default; inspect custom tools and settings too.

What is a good first task? A small fix with existing tests, so you can verify model access and the environment.

[ikaijua]

Official references

• Repository and installation
• Documentation
• Agent modes
• Desktop downloads