Configuration examples

These examples are organized by real file layout, not by stuffing multiple files into comments inside one YAML blob. Pick a scenario and create the files at the paths shown on that page.

Choose a starting point

Scenario	Best for	Page
Minimal	Smallest working setup: one provider, one key, one model pool	Minimal
Codex + LSP	Codex OAuth, Go / Python LSP, plus a reviewer agent	Codex + LSP
OpenAI-compatible gateway	Multi-key rotation and backup-endpoint failover	OpenAI-compatible gateway
Team repository	Project-level .chord/, hooks, shared commands, multi-agent roles	Team setup

These examples are starting points, not rigid templates. For field semantics and the full config surface, see the Configuration cheatsheet. If you already know the provider/model you want and only need a minimal copy-paste snippet, start with Model configuration recipes.

Agent file formats

Agent definitions can be written as .md, .yaml, or .yml files. The examples use .md because front matter keeps structured settings near a Markdown prompt body; use YAML files when you prefer a single plain YAML document. Both formats are loaded from the same global and project-level agents/ directories.

Context and output limits

The bundled example configs set limit.context and limit.output per model; examples include limit.input only when the provider publishes a separate input cap. For what each field means — and how Chord derives the input budget when a provider does not publish a separate input cap — see the Glossary. For how those limits interact with compaction, see Context management — Compaction.

Where things go

File	Path
Global config	~/.config/chord/config.yaml
Credentials	~/.config/chord/auth.yaml
Project-level overrides	<repo>/.chord/config.yaml
Global custom agents	~/.config/chord/agents/<name>.md
Project-level custom agents	<repo>/.chord/agents/<name>.md
Project-level slash commands	<repo>/.chord/commands/<name>.md

For the full layout and project-key resolution rules, see Paths.