Singularity vs Paradev (formerly Superset): agent sessions or a ticket board?

Paradev, formerly Superset, is built for developers who already use command-line agents such as Claude Code, Codex, OpenCode or Gemini CLI and want to run several at once without them stepping on each other. Its desktop app creates one workspace per task, backed by a Git worktree, and brings together terminals, diffs, ports and pull requests.

Singularity starts from the same principle (one agent, one isolated branch) but organises the work around a Kanban board where every ticket is an agent’s mission. So the real decision is not “which one runs more agents” but: do you want to steer agent sessions and GitHub pull requests, or a flow of tickets that land one by one on your base branch?

Name, plans and links updated on 27 October 2026; workflows checked on 10 October 2026 against Paradev’s official documentation and Singularity’s own docs. It compares workflows; it does not benchmark speed or the quality of generated code. Public prices in USD/EUR as displayed, excluding taxes and AI provider usage; they may change.

In short

• Same foundation: one Git worktree per task, and your own CLI agents with your own accounts. Neither resells a model subscription.
• Paradev centres on the session and the pull request: a very wide agent catalogue, a Tasks view, a coordinator agent, GitHub PR review and merge, and an iPhone/iPad app on Pro.
• Singularity centres on the ticket: dependencies, child tickets, a scheduler, a report, file-by-file review, rebase integration and a commit queue.
• Platforms and price: Paradev targets macOS and charges for team use; Singularity runs on macOS, Windows and Linux, and only its paid-plugin licence is billed.

What Paradev is

Paradev describes itself as a platform for running coding agents in parallel, “source-available” under the Elastic License 2.0 according to its documentation, and available as a desktop app, an iPhone and iPad app, a CLI and an MCP server. Its basic unit is the workspace: an isolated Git worktree on its own branch, with conflicts surfacing at merge time like for any branch. A deleted workspace keeps its worktree for 30 days (workspaces), and a .superset/config.json file defines setup, teardown and run scripts (scripts).

Paradev does not ship its own agents: it launches the ones you already have in its terminals. Its integration page lists more than twenty, including Claude Code, Codex, Gemini CLI, OpenCode, Copilot, Cursor Agent, Mistral Vibe, Amp and Kiro (agent integration); a built-in chat can also use your API keys through OpenRouter, AWS Bedrock or Google Vertex (providers).

The Tasks view offers a list and a status board, mixes native tasks with Linear issues, and its “Run in workspace” action creates one workspace per task with the agent already running (Tasks). An orchestration skill lets Claude Code or Codex coordinate workers, each in its own workspace, via superset agents create (orchestration). The diff viewer handles staging, committing and pushing (diff viewer); the Pull requests view sends a line comment to an agent and merges through GitHub with squash, merge commit or rebase (pull requests). ⌘O opens the workspace in Cursor, VS Code, Zed, a JetBrains IDE or other editors (IDE integration).

The app runs on macOS (Apple Silicon and Intel); a Linux x64 AppImage is experimental and Windows is not yet available (installation). Free costs nothing for one local user; Pro costs 20 USD per user per month, or 16 USD/month billed yearly (192 USD/user/year), and adds unlimited users, remote access, automations, Linear, Slack and mobile; Team is $60/user/month monthly or $54/month billed yearly ($648/user/year), adding beta cloud workspaces to Pro: 1,200 GB-hours per seat/month (150 hours on an 8 GB machine), 5 environments and $0.06 per extra GB-hour; Ultra is $200/user/month, billed monthly, with 4,000 GB-hours (500 hours) and 10 environments; Enterprise is custom-priced, with annual billing only (Paradev pricing). The team sets out its approach in a guide to parallel coding agents.

Cloud quotas are pooled across team seats and reset monthly. Advertised hours assume a 4 vCPU, 8 GB machine: doubling memory consumes the pool twice as fast. Extra GB-hours bought on demand cost $0.06 each and never expire. Free includes unlimited projects and local workspaces. Source: official pricing table and FAQ, checked on 27 October 2026.

Singularity in a nutshell

Singularity is a desktop ADE (Agentic Development Environment) from Meteor Factory: an editor with language servers, native terminals, a Git client (graph, diffs, stash, worktrees), multi-project workspaces and a Kanban board that is the execution surface for agents. A ticket carries the description, priority, dependencies and agent configuration, then keeps the report and a short resolution.

In worktree mode, each run works in its own Git worktree; commits are replayed onto the base by rebase, an integration queue orders tickets within a project, and the result is integrated automatically or waits in the Review tab. Singularity drives seven providers through their CLIs (Claude Code, Codex, Copilot, Cursor, Gemini, Mistral, Ollama) with your accounts. The app is free; the Singularity Licence (24 EUR per month per workstation, 14-day trial) only unlocks paid plugins in the Marketplace, such as Pipelines. It is proprietary software, available on macOS, Windows and Linux (installation).

Comparison table

The same feature in both tools

Take a common request: a CSV export of invoices, with an API endpoint and a button in the UI. The button depends on the endpoint.

In Paradev, you create two tasks or two workspaces; the setup script installs dependencies, then you might run Claude Code on the endpoint and Codex on the button. Since no formal dependency between tasks is documented, you (or a coordinator agent) decide when to start the button: right away with an API contract in the prompt, or after the endpoint is merged. Each workspace ends in a pull request that you read, comment on and merge.

In Singularity, the “button” ticket is marked Blocked by the “endpoint” ticket. The scheduler runs the endpoint, waits for its integration, then starts the button on a base that already contains the API. The card shows the agent’s phase; a blocking question puts the ticket on hold; the agent finishes with a report. With auto-merge off, you approve each file, annotate or relaunch with feedback, then integration happens by rebase.

Either way, quality depends mostly on the prompt and the tests. The difference is what the tool remembers: sessions, diffs and PRs in Paradev; a ticket, its dependencies and its result in Singularity.

Tracking, review and integration

Paradev thinks in workspaces and sessions. That is flexible: any CLI agent, several agents on the same problem to compare, or a coordinator splitting the work. The lasting record remains the branch, the PR and your issue tracker, Linear first. The end of the cycle runs through GitHub: line comments sent back to an agent, merges with the same effects as on github.com. For a team whose review already happens in PRs, that is a natural extension.

Singularity makes the ticket the record of the work: the board card, the brief the agent reads, and the thread that keeps attempts, exchanges and the report. Statuses separate an active run, an explicit wait and a retryable failure; a parent can wait for its child tickets, then synthesise their results. The Review tab lets you approve file by file, merge only approved files or relaunch with your notes; the commit queue serialises concurrent integrations. You then push and open a PR as usual. In both tools, a clean rebase or merge does not prove there is no regression: run the tests.

Cost, platforms and data

Both tools leave the model bill with your providers. Paradev is free for one local user and becomes paid for teams, remote access, automations or mobile. Singularity does not charge for the app, whatever the number of agents, projects or worktrees.

Paradev is a macOS app with experimental Linux, but offers an iPhone/iPad app that Singularity lacks; Singularity ships installers for macOS, Windows and Linux. Both run agents on your machine by default: Paradev says repositories and sessions stay on your machine, and Singularity keeps its data in a local SQLite database. Prompts and code excerpts go through the AI provider you choose, unless you use a local model through Ollama.

Where Paradev is stronger

• Agent catalogue: more than twenty documented CLI agents.
• GitHub flow: PR list, line comments sent to agents, merge from the app.
• Automation: CLI, TypeScript SDK, MCP server and a coordinator agent.
• Mobility and teams: iPhone/iPad app, remote access, Linear and Slack integrations on Pro.
• Licence: source code available under the Elastic License 2.0.

Where Singularity stands out

• The ticket as a mission: dependencies, child tickets, a scheduler and a concurrent-agent limit.
• Fine-grained review: per-file approval, notes on a selection, partial merge.
• Linear history: rebase and a commit queue, with no mandatory PR.
• Complete environment: editor, 31 available language servers, terminals and a Git cockpit.
• Platforms and price: macOS, Windows and Linux, and a free app.

Can you use both?

Yes, as long as two tools never manage the same branch. Both work on ordinary Git repositories and drive the same CLIs (Claude Code, Codex, Copilot, Cursor, Gemini, Mistral). You could keep Paradev for explorations and PR review, and Singularity for work split into dependent tickets. Git does not allow one branch in two worktrees: let each tool create its own branches.

Which tool should you choose?

• On a Mac, living in GitHub PRs, keen to try many agents: Paradev.
• Following agents from an iPhone or sharing access across a team: Paradev Pro.
• Work split into dependent tickets with a report per mission: Singularity.
• Windows or Linux machines: Singularity.
• Editor, Git and agents in a single app: Singularity.

For another Mac agent manager, see Singularity vs Conductor.

Trial protocol

1. Pick a repository with tests and a feature in two dependent parts.
2. Use the same agent (for example Claude Code) in both tools.
3. In Paradev, create two workspaces with a setup script; in Singularity, two tickets linked by “Blocked by”, with auto-merge off.
4. Count your interventions: questions, relaunches, handling the order.
5. Review, integrate, then run the full test suite on the base branch.
6. A week later, check what is still easy to find: PRs and sessions, or tickets and reports.

Frequently asked questions

Do Paradev or Singularity include a model subscription?

No. Paradev says you use your own accounts or keys on every plan, and Singularity drives your providers’ CLIs with your accounts. Usage is billed by those providers.

Does Paradev run on Windows?

Not as of 10 October 2026: its install page says Windows is not yet available and Linux x64 is experimental. Singularity ships macOS, Windows and Linux packages.

Is Paradev open source?

It describes itself as “source-available” under the Elastic License 2.0: the code can be read, with that licence’s restrictions. Singularity is proprietary.

Do worktrees prevent every conflict?

They stop two agents from writing to the same working files. Git or semantic conflicts can still appear at merge time in Paradev, or at rebase time in Singularity; tests and review remain essential.

To go further, browse all Singularity comparisons, including Singularity vs Claude Code, follow the getting started guide or download Singularity.