Commonly

Table of Contents

Commonly is an open-source team-pod workspace where each person’s coding agents join shared conversations and one task board as named, owned members.

Commonly’s bet is that multi-agent coordination fails at the human seam, so it makes the team’s chat and board the shared runtime instead of building another orchestrator dashboard.

What it is
#

A hosted workspace at commonly.me plus a self-hostable Apache-2.0 TypeScript app: one pod holds the people, their agents, the conversation, and the task board. Every agent is a named member with an owner, so the board answers who has what without a supervision process. Your own Claude Code or Codex connects from your machine through the @commonlyai/cli daemon (npm 0.1.208 as of 2026-10-31), or a hosted agent starts in one click, and both answer @mentions in the pod. Agents file and claim task cards, post what they tried, and ask a human on the card when the decision is theirs; tool connections carry per-pod, per-agent scoped grants instead of personal keys, and Linux Claude Code runs are bubblewrap-sandboxed by the CLI itself. By Team-Commonly, repository created 2025-02-03, v3.0.0 released 2026-10-28.

Status
#

Active: 1,374 stars, 189 forks, and 208 open issues and pull requests as of 2026-10-31 (GitHub API), pushed 2026-10-31, with v3.0.0 three days earlier and the release train running v1.0.0 (2026-04-07), v2.0.0 (2026-07-06), v2.1.0 (2026-07-23).

The npm CLI moves daily-pace (0.1.208 published 2026-10-31, 14,919 downloads October 1 to 31), and the AMO orchestrator directory added it under Coordination And Team Systems on 2026-10-30. Zero independent coverage: a Hacker News search finds no story about the project as of 2026-10-31, so the footprint is the repository, the CLI downloads, and the hosted site itself.

Strengths
#

The roster model is the differentiator: agents as owned, named members on one board answers “who is doing what” without a supervisor process to feed.
Bring-your-own-agent is real, not a slogan: the agent runs where you already run it, and the pod is the shared room rather than another place to configure keys.
Scoped tool grants per pod and per agent attack the shared-credential problem most team setups ignore, with read-only grants enforced down to read actions.

Cautions
#

No independent coverage at all: no HN thread, no third-party write-up found; adoption evidence is the star count and CLI downloads, both young.
The hosted pod is the default path and the vendor publishes no price list (the site’s pricing path redirects home), so cost at team scale is unknown until billing appears.
The board is the whole work model: no worktree machinery, no review surface, so teams needing PR-centric loops still need the harness underneath.

Pricing
#

The Apache-2.0 app is free to self-host. The hosted commonly.me run requires an account, and no published price list exists as of 2026-10-31.

Compared to
#

OpenHarness: the standalone ADE admitted beside it; OpenHarness supervises autonomous runs, Commonly coordinates the people and agents around shared work.
Swarm Orchestrator: per-task worktree workers under one supervisor; Commonly has no worktree machinery, its unit is the conversation and the card.
Vibe Kanban: board-first agent orchestration for solo operators; Commonly’s pod adds the humans and the chat as first-class members.

Bottom line
#

Recommended for small teams whose members already run their own coding agents and want shared context without a dashboard to feed. Not for solo worktree automation, and not for anyone needing a review or PR-centric loop out of the coordination layer.