Trying a new agent without losing your history

Many agents, one playbookDeveloper testing OpenCode

We wanted to try OpenCode with an open model, but not at the cost of everything the playbook had learned.

How can I try OpenCode with an open model without losing my project's history?

Run arcroom init --agent opencode to connect OpenCode to the same ArcRoom project memory. From its first prompt it gets the decisions, rules and saved answers your earlier agents built up. If it doesn't suit you, switch back and nothing is lost. OpenCode support is experimental.

Also asked as: I want to test a new AI agent but don't want to lose what my current one learned. How? · How do I give OpenCode the same project memory as my other agents? · Can OpenCode see decisions my other AI agents saved?

The work

arcroom init --agent opencode connects it to the same project memory.

ArcRoom keeps
  • Decision
The next agent gets

[opencode] Team decisions (3) and code map loaded for this project.

Step by step

  1. arcroom init --agent opencode connects it to the same project memory.
  2. The new agent gets decisions, rules and saved answers from its first prompt.
  3. If it doesn't suit you, switch back. Nothing is lost.

What the agent sees

[opencode] Team decisions (3) and code map loaded for this project.

Compared with today's tools

AGENTS.md copied from CLAUDE.md

Copy the team's CLAUDE.md rules into AGENTS.md so OpenCode starts with the same rules.

Setup
10 min
Ongoing
keep two files in sync

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min (npx arcroom init --agent opencode, experimental)
Ongoing
None

Rules carry over, but saved answers, decisions and last-run notes from Claude sessions don't.

Where it stops

OpenCode support is experimental.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in many agents, one playbook