# Run OpenCode in parallel — Tallos

> Tallos runs OpenCode, the open-source terminal coding agent that works with many model providers, as parallel sessions on macOS and Windows. Each session gets its own git worktree, and Tallos lists OpenCode's models so you can pick one per session or per squad member. OpenCode keeps using the provider logins you already set up.

- Canonical: https://runtallos.com/integrations/opencode
- Português: https://runtallos.com/pt/integracoes/opencode.md
- Updated: 2026-09-29
- Section: Integrations

## Key facts

- Install and sign in from Settings → Connections
- Models listed live from your OpenCode install
- One isolated git worktree per session
- Pick a model per squad member, run several copies

## What is OpenCode?

**OpenCode** is an open-source AI coding agent built for the terminal and run with the `opencode` command. It is provider-agnostic: you connect the model providers you want and pick a model per session. Like other agent CLIs, it reads your code, edits files and runs commands.

**Tallos** is a desktop [agent development environment](https://runtallos.com/what-is-an-ade) that launches the real OpenCode CLI in isolated workspaces, alongside Claude Code, Codex, Gemini CLI and your other agents.

## How do I set up OpenCode in Tallos?

Open **Settings → Connections**, install OpenCode from its card, sign in to a provider, and set it as your default agent. If `opencode` is already on your PATH, Tallos detects it.

1. **Install** — Click **Install**. Tallos shows the official command before running it: `brew install anomalyco/tap/opencode` or `npm install -g opencode-ai@latest` on macOS, and `npm install -g opencode-ai@latest` on Windows (Node.js required).
2. **Sign in to a provider** — Click **Sign in** to run `opencode auth login` in a terminal. OpenCode stores each provider's credentials itself, so Tallos does not ask for an API key for it. The card shows **Connected** once OpenCode has saved a login.
3. **Check the models** — Open **Models** on the card. Tallos asks your OpenCode install which models it can use, so the list matches the providers you connected.
4. **Make it the default and start** — In **Settings → Agents**, set OpenCode as the **Default agent**. Then add a repository, click **+** for a workspace and describe the task; Tallos launches OpenCode in a fresh worktree with your prompt.

For every agent on the Connections page, see [agent connections](https://runtallos.com/features/agent-connections).

## How do I run multiple OpenCode sessions in parallel?

Open one workspace per task. Each workspace is a separate git worktree with its own branch, files and terminals, so several OpenCode sessions can work on one repository at the same time without touching each other's files.

- **Different models, same task.** Run one session per model and compare the diffs side by side.
- **Status at a glance.** Tallos installs a small status plugin so the sidebar shows when each OpenCode session is working or waiting for you.
- **Review per branch.** Each workspace has its own diff; comment on lines and send the feedback back to that session.
- **Resume later.** OpenCode sessions show up in Agent Session History; **Resume** runs `opencode --session <id>` in the right folder.

The pattern behind this is covered in [run AI agents in parallel](https://runtallos.com/run-ai-agents-in-parallel) and [parallel workspaces](https://runtallos.com/features/parallel-workspaces).

> Run OpenCode across as many workspaces as the task needs. → https://runtallos.com/signup

## How do I use OpenCode alongside other agents?

Add it to a [squad](https://runtallos.com/features/squads): a leader agent plus members, where each slot picks its own agent and model. OpenCode's model list is fetched live, so a squad member can use any model your OpenCode install supports.

- **Several copies of one member.** Set **How many** to 4 and the leader gives each OpenCode copy its own module — the [test coverage squad](https://runtallos.com/squads/test-coverage) works this way.
- **Reviewer or tester.** Pair a builder with an OpenCode member that reviews the fix or writes the regression test.
- **One of three attempts.** In *3 attempts, pick the best*, OpenCode can be one of the agents trying the same objective.

Squads run inside the limits you set — 5 rounds and 4 members at once by default, up to 20 rounds and 12 members. Every member works in its own worktree, and nothing merges without your approval.

## Does OpenCode open in a chat view?

No. OpenCode opens in the terminal view, with its own interface exactly as you know it. The chat view is available for agents whose transcripts Tallos can read, such as Claude Code, Codex and Grok CLI.

Tallos waits for OpenCode's input box to be ready before it pastes your prompt, so the task text is not lost during startup.

## What does Tallos add to OpenCode?

| Capability | OpenCode on its own | OpenCode in Tallos |
| --- | --- | --- |
| Parallel sessions | Open more terminals and set up git worktrees yourself | One workspace per task, each with its own worktree and branch |
| Install and sign-in | Run the install and `opencode auth login` yourself | Install and sign in from Settings → Connections |
| Model choice | `opencode models`, then pass `--model` | Models listed on the card; pick one per session or squad member |
| Session status | Visible in its own terminal | Working / waiting across all sessions in the sidebar |
| Past sessions | `opencode --session` from the right folder | Searchable history across agents, one-click Resume |
| Other agents | Separate tools | Same app; squads with a leader, members and copies |

_OpenCode stays provider-agnostic. Tallos adds the workspaces and coordination._

## Frequently asked questions

### Do I need a separate subscription to use OpenCode in Tallos?

No. OpenCode keeps using the provider accounts and keys you connected with `opencode auth login`, and Tallos does not resell model access.

### Why can't I paste an API key for OpenCode in Tallos?

OpenCode works with many providers and stores each provider's key itself through `opencode auth login`. Tallos opens that flow instead of keeping a separate key.

### Can I run OpenCode with different models at the same time?

Yes. Start one workspace per model, or add several OpenCode members to a squad with a different model each. Every session has its own worktree.

### Does Tallos change OpenCode's permission settings?

No. Tallos passes no permission-bypass flag to OpenCode, so it follows the permissions in your OpenCode configuration.

### Does OpenCode work in Tallos on Windows?

Yes. On Windows 10+, the Connections page installs OpenCode with npm, so install Node.js first.

### Can OpenCode lead a squad?

Yes. The leader can be any supported agent, including OpenCode. It plans the work, hands out tasks with the `tallos` command and reports back to you.

---

**Run your agents in parallel with Tallos** — Claude Code, Codex, Gemini and 30+ agents — each in its own workspace, with squads, chat, terminals and review in one app. Uses the subscriptions you already have.

https://runtallos.com/signup (macOS 13+ · Windows 10+)