Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 18 additions & 4 deletions apps/marketing/content/category/agent-orchestration.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
title: "AI Agent Orchestration for Software Development"
description: "What agent orchestration means for coding: the primitives (isolation, sessions, review, scheduling, programmatic control), the patterns, and when you actually need it."
date: 2026-08-17
lastUpdated: 2026-08-17
lastUpdated: 2026-09-23
keywords:
- agent orchestration
- ai agent orchestration
Expand Down Expand Up @@ -30,7 +30,7 @@ You do not need orchestration to run one agent, and you barely need it for two.
- More than one person (or agent) needs to see what the fleet is doing.
- Losing a session's state (a laptop sleep, an app restart) costs real work.

If your agents stop the moment you close your laptop lid, you have sessions, not orchestration.
A local machine must remain awake for its agents to keep executing. To continue work while your laptop is closed, run it on an available remote host.

## The primitives

Expand All @@ -42,7 +42,7 @@ Each unit of work gets its own working directory and branch, almost always a [Gi

### Persistent sessions

Agent sessions need to survive the app restarting, the terminal closing, and the machine sleeping. Without persistence, orchestration degrades into re-prompting. In Superset this is a terminal daemon that owns the sessions; the app is just a view onto them.
Session persistence separates the agent process from the app window. In Superset, a terminal daemon owns the sessions, so restarting the app does not need to restart the agent. Persistence does not keep a sleeping or powered-off host executing work. See the [terminal daemon implementation](/blog/terminal-daemon-deep-dive).

### Status and monitoring

Expand All @@ -66,6 +66,20 @@ An API for the whole loop, so scripts and other agents can create workspaces, la

![The superset CLI menu: agents, automations, hosts, projects, tasks, terminals, and workspaces, all scriptable from any shell](/changelog/2026-08-09-cli-browser.png)

## What Superset orchestrates

Superset (superset.sh) is a source-available workspace for CLI coding agents. It is unrelated to Apache Superset, the business-intelligence product. Claude Code, Codex, and OpenCode provide the coding agents; Superset provides workspaces, terminals, review, and scheduling around them.

| Fact | Where to verify it |
|---|---|
| Source-available under Elastic License 2.0, not OSI-approved open source | [Repository license](https://github.com/superset-sh/superset/blob/main/LICENSE.md) |
| macOS desktop; experimental Linux AppImage; Windows unavailable | [Downloads](/download) |
| Free tier plus paid seats; agent-provider usage is separate | [Current plans](/pricing) |
| Automations require Pro or Enterprise and an available target host | [Automation requirements](https://docs.superset.sh/automations) |
| Worktrees separate working files, not process permissions, ports, or databases | [Git worktree reference](https://git-scm.com/docs/git-worktree) |

For a first scheduled task, use the [maintenance-check walkthrough](/blog/scheduled-agent-maintenance). It includes a bounded prompt and explains how to inspect the result. An automation run marked **created** means its workspace was created; inspect the agent output to know whether the checks passed.

## Common patterns

**Fan-out.** Split a large task into independent subtasks, launch an agent per subtask, review the diffs as they land. The default pattern, and the reason isolation is non-negotiable.
Expand Down Expand Up @@ -114,7 +128,7 @@ Yes. Expose orchestration as MCP tools and any MCP-capable agent can create work

### How do scheduled agent runs work?

You define a prompt, an agent, and a schedule; the orchestrator creates a fresh workspace at the appointed time, runs the agent, and leaves the result for review. Superset calls these automations. Typical uses: nightly dependency bumps, issue triage, recurring cleanup passes.
You define a prompt, an agent, a target host, and a schedule. Superset Automations require Pro or Enterprise. If the host is offline when dispatch occurs, the run fails and the schedule advances; retry after the host is available. A created workspace is not evidence that the agent succeeded. Read its output before accepting the result.

### What should I look for in an agent orchestration tool?

Expand Down
29 changes: 26 additions & 3 deletions apps/marketing/content/category/parallel-coding-agents.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
title: "Parallel Coding Agents: The Complete Guide"
description: "What parallel coding agents are, why one agent session stops being enough, and how to run many agents on one repository without losing track of any of them."
date: 2026-08-17
lastUpdated: 2026-08-17
lastUpdated: 2026-09-23
keywords:
- parallel coding agents
- parallel ai coding agents
Expand Down Expand Up @@ -86,6 +86,29 @@ Three kinds of tools support parallel coding agents today, and they compose rath

For a full comparison across all three layers, see [Best AI Coding Tools and Agents](/compare/best-ai-coding-agents-2026).

## A reproducible isolation check

We checked this on September 23, 2026 with Git 2.52.0 in a temporary repository: two task branches changed the same line independently. The main checkout stayed unchanged. After merging the first branch, merging the second returned a conflict in `README.md`. No coding agent was involved in this check; it tests Git isolation, not agent performance.

To repeat the check in a disposable repository, commit a one-line `README.md`, then create two worktrees:

```bash
git worktree add -b task-a ../task-a
git worktree add -b task-b ../task-b
```

Change the same line to different text in each worktree and commit each change on its own branch. In the original checkout, run:

```bash
git merge --ff-only task-a
git merge --no-edit task-b
git diff --name-only --diff-filter=U
```

The last command should list `README.md`. Use `git merge --abort` to leave that deliberate conflict. The [Git reference](https://git-scm.com/docs/git-worktree) explains which repository state worktrees share.

For real tasks, give dev servers separate ports and use separate test data where needed. Review and test the combined branch after integration, even when Git merges it without a conflict. The [three-task walkthrough](/blog/parallel-coding-agents-guide) covers prompts, previews, and review; the [Claude Code guide](/compare/multiple-claude-code-agents-parallel) covers session setup.

## Where Superset fits

Superset is a workspace built specifically for this workflow: every task gets an isolated worktree with a persistent terminal, any CLI agent works, and monitoring, diff review, port handling, automations, and an MCP server for programmatic control sit on top. If the problems in this guide are the ones you are hitting, that is exactly what it exists for. [Download it](/download) or read [how it compares](/compare) to everything else in the space.
Expand All @@ -94,15 +117,15 @@ Superset is a workspace built specifically for this workflow: every task gets an

### What are parallel coding agents?

Running more than one AI coding agent at the same time, each on its own task, typically with each agent isolated in its own Git worktree and branch so their changes never collide. The point is throughput: you review and direct while agents work concurrently.
Running more than one AI coding agent at the same time, each on its own task, typically with each agent in its own Git worktree and branch so they edit separate working files. Changes can still conflict when merged. The point is throughput: you review and direct while agents work concurrently.

### Do I need special tooling to run agents in parallel?

Not to start. Two terminal tabs and two Git worktrees are enough to try it. Tooling becomes worth it when you hit the predictable problems: losing track of sessions, reviewing many diffs, sessions dying on sleep, and worktree cleanup. That is the point of orchestration workspaces like Superset.

### How do parallel agents avoid conflicts?

Isolation. Each agent works in its own Git worktree on its own branch, so edits never touch another agent's files. Conflicts only surface at merge time, where Git handles them the same way it handles human branches.
Give each agent its own working directory and branch, and assign tasks with limited overlap. Worktrees do not restrict an agent's filesystem permissions or isolate shared services. Merge conflicts and behavioral conflicts can still occur, so review the combined diff and run the relevant tests after integration.

### How many coding agents can I run at once?

Expand Down
Loading
Loading