{fleet}
Weave
v0.13.0 · MIT License

Orchestrate a fleet of
AI coding agents.

One web UI. Multiple workspaces. Parallel execution.

Spawn, monitor, and interact with multiple OpenCode sessions from a single browser tab. Each agent works independently — different repos, different tasks, zero context collision.

Install Now View on GitHub
Fleet web dashboard showing multiple OpenCode sessions running in parallel with status indicators, message streams, and diff viewers
Built for the OpenCode ecosystem
{install}

Install in One Command

curl -fsSL https://get.tryweave.io/agent-fleet.sh | sh
2

Then run:

weave-fleet

Managing your installation

  • weave-fleet update — Update to the latest version
  • weave-fleet version — Check installed version
  • weave-fleet uninstall — Remove Fleet

Requires OpenCode CLI · Full setup guide

{features}

Features

Everything you need to run a fleet of AI coding agents

🖥️

Multi-Session Management

Spawn and monitor multiple OpenCode sessions. Each runs in its own workspace with independent context.

📊

Web Dashboard

Real-time web UI on localhost:3000. See all sessions, their status, diffs, and event streams at a glance.

🔒

Workspace Isolation

Each agent session gets its own workspace. No context bleeding between tasks. True parallel development.

🎮

Session Controls

Create, prompt, abort, and inspect sessions. Send follow-up prompts. View generated diffs inline.

⚡

One-Command Install

Single curl or PowerShell command. Self-contained binary with bundled Node.js. No dependency management.

🔄

Auto-Updates

Run weave-fleet update to pull the latest version. Always current, zero friction.

🌐

Cross-Platform

macOS, Linux, and Windows. Same experience everywhere.

💾

SQLite Persistence

Session state persisted to ~/.weave/fleet.db. Restart without losing context.

{how_it_works}

How It Works

From install to orchestration in under a minute

1

Setup

Install

Run the one-liner for your platform. The installer downloads the self-contained bundle — including a Node.js runtime — and creates the weave-fleet command.

$ curl -fsSL https://get.tryweave.io/agent-fleet.sh | sh
✓ Installed to ~/.weave/fleet
↓
2

Start

Launch

Run weave-fleet and open localhost:3000. The dashboard shows your workspace and lets you create new agent sessions.

$ weave-fleet
✓ Fleet running at http://localhost:3000
↓
3

Use

Orchestrate

Spawn agents across repos. Monitor progress, view diffs, send prompts — all from one tab. Agents work in parallel, each in isolation.

● session-1 api-refactor
● session-2 auth-module
● session-3 test-suite
3 agents running in parallel
{orchestration_skill}

Unlock Agent-Driven Orchestration

Fleet isn't just a dashboard for you to manage sessions. Install the fleet-orchestration skill, and your agents manage sessions themselves.

🧠

Autonomous Parallelization

Agents analyze task overlap and decide whether to parallelize or serialize — before spawning any children. No manual coordination needed.

🔀

Agents Spawn Agents

A parent agent creates child sessions, assigns scoped tasks, and sets file boundaries — each child works in its own isolated worktree.

📡

Callback-Driven Workflow

No polling. When a child finishes, Fleet sends a callback to the parent with a summary of files changed and final status. The parent inspects results and moves on.

⚠️

Automatic Conflict Detection

After parallel children complete, the parent reviews all diffs for overlapping file changes and flags conflicts before reporting success.

🔄

Error Recovery

Failed child sessions are automatically inspected. The parent can retry with clarified instructions or escalate to the user for guidance.

⚡

One Skill Install Away

Install the fleet-orchestration skill from Settings, assign it to Loom and Tapestry, and your agents gain the ability to orchestrate multi-session workflows.

Read the Skills setup guide →
{fleet_and_weave}

Fleet + Weave

Two products. Complementary. Independent.

Fleet works with plain OpenCode

Fleet manages multiple OpenCode sessions — no plugins required. Each session runs standard OpenCode. The Weave plugin is completely optional.

Weave enhances each session

If you DO install Weave, each Fleet session gets Weave's structured Plan → Review → Execute → Audit workflow with automatic code quality and security gates. Fleet manages the sessions; Weave enhances what happens inside them.

Different layers of orchestration

Fleet = multi-session management (across repos, parallel tasks). Weave = multi-agent orchestration (within a single session). They complement each other but are fully independent.

{config}

Configuration

Configure via environment variables — sensible defaults out of the box

Variable Default Description
PORT 3000 Server port
WEAVE_HOSTNAME 0.0.0.0 Server bind address
WEAVE_DB_PATH ~/.weave/fleet.db SQLite database path
WEAVE_INSTALL_DIR ~/.weave/fleet Installation directory
{faq}

Frequently Asked Questions

Do I need the Weave plugin?

No. Fleet works with plain OpenCode — no plugins required. If you install Weave, each session gets enhanced multi-agent orchestration, but it's entirely optional.

How many sessions can I run?

Limited by system resources. Each session runs an OpenCode server process. On a typical dev machine, 5–10 concurrent sessions is comfortable. Monitor memory and CPU usage.

Is Fleet free?

Yes. MIT licensed, open source, free forever. You pay only for the AI model API calls each session makes.

Is my data stored locally?

Yes. All state is persisted to SQLite at ~/.weave/fleet.db. Nothing leaves your machine. No telemetry, no external APIs.

What platforms are supported?

macOS, Linux, and Windows. Fleet ships as a self-contained bundle with its own Node.js runtime — no system dependencies beyond OpenCode.