Overview

Features are the tools you drop into a workbench: AI coding agents, but also plain CLIs and SDKs. They land as devcontainer features - you pick them explicitly in the yml, and Monoceros installs them, pre-authenticated, into the container. Not everything here is an AI tool; the GitHub and GitLab CLIs are ordinary command-line tools, for example. This page covers the shared model; each feature has its own page with details.

New to Monoceros? Start with Installation.

The catalog

Catalog name	Tool	Kind
atlassian	Atlassian Rovo Dev (acli) + Teamwork Graph (twg) + Forge (forge)	AI + CLI
claude	Anthropic Claude Code	AI agent
claude-code-roles	Plan / implement / review roles for Claude Code	Workflow
github	GitHub CLI (gh)	CLI
gitlab	GitLab CLI (glab)	CLI
graphify	Graphify: the workspace as a queryable knowledge graph	AI + CLI
opencode	sst OpenCode (provider-agnostic)	AI agent
opencode-roles	Plan / implement / review roles for OpenCode	Workflow

The Atlassian feature has sub-variants: atlassian enables all three tools, atlassian/twg only the Teamwork Graph CLI, atlassian/rovodev only Rovo Dev, atlassian/forge only the Forge CLI. Beyond the catalog, any devcontainer feature works - pass its full OCI ref to add-feature.

Adding a feature

At init, when you create the container (comma-separated for several):

monoceros init acme --with-features=claude,github,atlassian/twg

Or later, on an existing one:

monoceros add-feature acme github

Both take a catalog name, or a full OCI ref for any other devcontainer feature.

Note

acme is just a sample workbench name - it’s whatever you passed to monoceros init. Substitute your own throughout.

Removing a feature

monoceros remove-feature acme github

Or delete the feature’s entry from acme.yml by hand and re-run monoceros apply acme - the yml is the source of truth.

Credentials

API keys and tokens go on the feature entry in the yml, as ${VAR} references; the real values live in acme.env (see Workbench configuration).

acme.yml

features:
  - ref: ghcr.io/getmonoceros/monoceros-features/claude-code:1
    options:
      apiKey: ${CLAUDE_CODE_API_KEY}

acme.env

CLAUDE_CODE_API_KEY=sk-ant-...

An optional credential can be left empty - for Claude Code, an empty apiKey falls back to OAuth / subscription login.

Tip

Reuse the same credentials across workbenches by putting per-feature defaults in monoceros-config.yml; apply merges the container yml over those defaults.

Note

The container explains itself: apply writes an AGENTS.md / CLAUDE.md briefing at the workspace root, so an AI tool knows the real stack (languages, services, tools) without per-session setup.