Home

jig Wiki

A terminal AI coding harness written in Go, in the spirit of OpenCode, built on the Charm stack. Run agents with tools against your codebase in an interactive TUI (jig) or headless (jig run).

Source: github.com/gammons/jig

Getting started

1. Installation — go install, build from source, first run.
2. Configuration — config.toml files, how they merge, substitution tokens, data paths.
3. Models and Effort — picking models, aliases, per-agent models, reasoning effort.

Using jig

• Usage — the TUI, jig run, and the other subcommands; flags and exit codes.
• Keybindings — modes, keys, the ctrl+p picker, remapping, themes.
• Agents — built-in agents, TOML and markdown agent definitions.
• Skills — SKILL.md discovery, using superpowers.
• Permissions and Trust — allow/ask/deny rules, subagent limits, project trust.
• MCP Servers — stdio/HTTP/SSE servers, .mcp.json, OAuth sign-in.
• Importing from opencode — bring opencode sessions into jig.

Reference

• Tools and Output Limits — the built-in tools and how their output is capped.
• Debugging — JIG_DEBUG and the debug log.
• Architecture — layers, and where the detailed notes live.

Project

• License: MIT