Using Era Code with OpenCode

OpenCode is the original Era Code harness. Era Code supports two AI coding harnesses — OpenCode (launched with era-code start) and Claude Code (launched with era-code claude). Both run the same governance, agents, commands, skills, and MCP integrations from a single template source. OpenCode is the canonical format that source is authored in; the Claude Code tree is translated from it.

Global Setup vs Per-Repo Init

There are two ways to wire Era into OpenCode, and they layer rather than compete:

• era-code setup (recommended, run once per machine) installs the full Era baseline at user scope — agents, skills, commands, MCP servers, permissions, governance, and the constitution + universal directives — into ~/.config/opencode/, plus global governance memory at ~/.era/memory/. After this, every repo inherits Era with no per-repo step. See era-code setup.
• era-code init (optional, per repo) writes a project .era/ directory with project-specific directives and syncs the project surface (AGENTS.md/opencode.jsonc, .opencode/).

Run setup once for the org-wide baseline; run init only when a repo needs its own directives.

What Gets Installed Where

Global resources

Every era-code command keeps ~/.era/era-code/ up to date, gated on the CLI version (a no-op when the installed version matches). The OpenCode tree is the canonical template source:

~/.era/era-code/opencode/
├── agents/        # primary agents + subagents (incl. personas)
├── command/       # slash commands
├── skill/         # skills (SKILL.md + references/)
└── _shared/       # shared context files (ethos, output compression, Linear workflow, ...)

Project files

era-code start syncs these project files on every launch:

Path	Ownership	Contents
.opencode/agent, .opencode/command, .opencode/skill	Era	Agents, commands, and skills for the project
opencode.jsonc	Era	OpenCode config: mcp: servers, permission: block, and instructions: loading the shared context files
AGENTS.md	Shared	An Era-managed marker block importing the constitution and directives; everything outside the markers is yours
DIRECTIVES.md	Era	Compiled from your directive files on launch

Configuration Files

• opencode.jsonc — OpenCode's config file. Era writes the mcp: block (servers with {env:VAR} credential interpolation; disabled servers are kept with enabled: false), the permission: block, and the instructions: list that loads the shared context files.
• AGENTS.md — the OpenCode equivalent of CLAUDE.md: an Era-managed marker block that imports the constitution and directives so governance travels with the repo.

Governance

OpenCode governance runs as the in-process era-governance plugin. It applies the same rule engine the Claude Code harness wires through a hook: hardcoded safety denials (privilege escalation, system destruction, secret extraction, private keys), default GitOps/IaC denials for kubectl/helm/gcloud/docker/gh/git, .era/governance.yaml overrides, and compound-command tokenization. The plugin no-ops when the project has no .era/ directory, and supports audit_mode deny logging.

Additional OpenCode-only plugins reinforce behavior at runtime:

• era-compression reinforces the output-compression doctrine per turn.
• era-auth refreshes auth tokens per command via a shell.env hook.
• era-handover tracks context-window usage and drives a two-phase handover.
• rtk-optimizer rewrites bash commands through RTK to compress shell output.

Agents

OpenCode exposes Era's primary agents — orchestration and plan — which are selectable and Tab-cyclable, with push-agent as a third primary. Personas auto-register as primary agents when the project's git remote matches a registered Era persona. Subagents are invoked automatically via the Task tool. Set the default primary agent with era-code config --set settings.default_agent=<agent>; see era-code start for the agent list.

Baseline MCP Servers

Linear, Notion Docs Reader, and Slack are baseline servers — the connections every Era engineer is expected to have. They are shown checked in the era-code init checkbox, force-re-added to existing manifests on CLI upgrade, and gracefully disabled with a one-line warning at launch when the credential is missing. Provide the token (re-run era-code init or export the env var) and the next launch re-enables it.

See the MCP Servers guide for Linear, Notion, and Slack setup.

Choosing a Harness

Both harnesses are first-class and share one template source, so governed behavior is equivalent. Practical differences:

• OpenCode offers Era's primary agents (orchestration, plan) with Tab cycling and plugin-enforced governance, and authenticates with Claude Max through the Meridian proxy.
• Claude Code authenticates with Claude Max natively (no proxy) and brings its own main agent and plan mode.

You can switch freely — era-code start and era-code claude each sync their own harness config from the same .era/ source of truth.