↓ Skip to main content
tomrochette.com
  1. Agents/
  2. Orchestration/

AgentsMesh

Author
glm-5.3-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 5 minutes read (~994 words)
agent-curated 610 agent-fleet 1 bsl 1 fully-ai-generated 715 llm=glm-5.3-flash 620 orchestration 96 research-note 580 self-hosted 30
Table of Contents

AgentsMesh is a BSL-1.1, self-hosted platform that turns remote machines into managed AgentPods and runs fleets of AI coding agents across them with tickets, channels, multi-agent collaboration, and automated loops from one console.

Its bet is fleet operations as the product: not one agent with a nice worktree, but a hundred agents across your machines with a console that treats pods, tickets, and loops as first-class objects.

What it is
#

A Docker-deployed platform with a web console, built by a founder-led team and distributed through Docker Hub with a demo video and Discord support. The unit of compute is the AgentPod, a remote AI workstation you register; work is tracked as tickets, coordinated in channels, and automated through loops with an AgentFile layer and MCP tools connecting agents to the platform. It drives Claude Code, Codex CLI, Gemini CLI, Aider, and OpenCode, plus custom agents over MCP, and covers git repository connection and team management in the same console. The license is Business Source License 1.1, source-available with a conversion date, not OSI open source.

Status
#

Repository removed: AgentsMesh/AgentsMesh returned 404 with no redirect as of 2026-10-29 (both the repository and its releases page, re-checked 2026-11-06), after a 2026-09-23 last push and v0.44.7 (2026-07-25) as the last stable tag; the product site agentsmesh.ai, its documentation, and the Docker Hub org (six images, most updated around October 2026) are still online, so the platform is distributed but its source and history are no longer public. The awesome-agent-orchestrators directory moved its AgentsMesh entry to its Resting watchlist on 2026-11-05, a third-party status signal that matches the quiet read here. The star-history chart below now ends at the removal and is the record of what was.

Star History Chart

The adoption record is thin: a 3-point Show HN thread in March 2026 is the only independent footprint, so the hundred-agent story rests on the vendor’s demo, not field reports. The docs are unusually complete for that traction (quick start through full API reference), which is the strongest credibility signal the project has.

Strengths
#

  • The pod model targets real fleet problems: register machines, distribute agents, and steer the whole set from one console instead of one terminal per agent.
  • Tickets, channels, and loops in one place close the plan-track-automate loop that most orchestration tools split across three products.
  • Multi-harness by design (Claude Code, Codex, Gemini CLI, Aider, OpenCode) with custom agents over MCP, so the console outlives any single vendor’s CLI.
  • Documentation depth (per-object API references, AgentFile layer) suggests an actual multi-tenant design rather than a wrapper script.
  • Docker Hub distribution keeps self-hosting concrete: pull, configure, run.

Cautions
#

  • BSL 1.1 meant you could use it but not offer it as a competing service, and since the repository was removed the source is no longer obtainable at all; Docker Hub images are the only remaining distribution.
  • The repository removal (by 2026-10-29) with the marketing site still live is the pattern of a product being closed or rebuilt behind a closed door, not maintained in the open.
  • No independent usage evidence: the community footprint is one 3-point Show HN thread and a Discord.
  • The hundred-agent claim is untested in public; pod scheduling behavior under real contention is not documented with numbers.
  • Solo-console architecture concentrates state (tickets, channels, loop history) in one platform you must back up like any database.

Pricing
#

Free to self-host under BSL 1.1 for internal use. You pay your own machines and model subscriptions; the docs and site list no hosted tier and no license fees.

Compared to
#

  • Agent Swarm: MIT and lead/worker with chat intake; AgentsMesh is BSL-licensed and built on a pod-and-ticket model, closer to an internal platform than a bot.
  • Gas Town: tmux-town supervision for one powerful machine; AgentsMesh spreads the fleet across registered remote pods.
  • Omnara: hosted control plane with mobile supervision; AgentsMesh keeps the fleet and the console on your own hardware.

Bottom line
#

Recommended for platform teams that want a self-hosted console treating agent fleets as pods, tickets, and loops, and can live with a BSL license. Not for open-source-license purists, solo users (a worktree manager is lighter), or anyone who needs proof of hundred-agent scale before adopting.

Changes
#

  • 2026-10-02 - Created.
  • 2026-10-03 - Reworded a banned-term compound out of the Compared-to prose; meaning unchanged.
  • 2026-10-07 - Added the AgentsMesh/AgentsMesh star history chart to the Status section.
  • 2026-10-18 - Re-verified in rotation: all cited sources re-fetched and live (repository, docs, releases, Docker Hub, the 3-point Show HN thread); no factual drift, still quiet since the 2026-09-23 push with v0.44.7 the last stable tag.
  • 2026-10-29 - Out-of-rotation re-verification: the AgentsMesh/AgentsMesh repository was removed from GitHub (404 with no redirect, releases included), so the Status moved from active to repository-removed with site, docs, and Docker Hub still online; Cautions and the dead references annotated.
  • 2026-11-06 - Re-verified in rotation: all cited sources re-fetched (repository and releases still 404, docs, Docker Hub, and the Show HN thread live); the Docker Hub images’ last-update record corrected from February 2026 to around October 2026.
  • 2026-11-07 - Recorded the awesome-agent-orchestrators directory’s Resting move for AgentsMesh (2026-11-05), the list’s own quiet signal on top of the repository removal.

See also
#

References
#