Thrum Introduction
Thrum lets the coding agents you already run, on Claude Code, Codex, Cursor and other tools, message each other and work as a team. You talk to one coordinator agent, which hands out the work, creates a reviewer on a different model when you ask, checks each branch and does the merging. Before it pushes a merge, it is set up to build and test the merged result, and an optional gate agent runs your tests and reports what it did and did not measure. Agents that write code are told to start from a failing test. You can message agents from an inbox, watch them in a web UI, and choose whether the coordinator asks you before each merge. Thrum runs locally and connects over git, LAN or Tailscale, and its maker says the only data he sees is for licensing. The Free version needs no account; Pro, at $15 a month or $120 a year, adds typing to agents from the web UI, agent memory and a work queue, and is free for registered public repositories. It was built by Leon, who started it as an open-source project in late 2025, and launched on Hacker News in November 2026.
Thrum key facts
| Launched | Nov 9, 2026 on Hacker News |
|---|---|
| Pricing | Freemium, from $15/mo |
| Free tier | Yes, Free needs no account |
| Platforms | Local install with a web UI |
| Sign-up | Not needed for Free |
| Made by | Leon, an independent developer |
How we checked Thrum
Not tested hands-on · Checked on Nov 10, 2026
Thrum is installed locally and works with your own coding agents, so we did not install it. We read the website and docs, and the maker's Hacker News launch post and replies.
Thrum Key Features
Agents Message Each Other
Agents on different tools talk directly, so you no longer paste one agent's answer into another's window.
Coordinator and Reviewer
One coordinator assigns the work and merges it, while a reviewer on a different model checks each branch.
Tested Merges
The coordinator builds and tests the merged result before pushing, and a gate agent can run your tests and report what it checked.
Runs Locally
Thrum connects your agents over git, LAN or Tailscale instead of a cloud service.
Thrum Use Cases
Developers running several coding agents: stop relaying messages between Claude Code, Codex and Cursor by hand.
Teams that want checked work: have a second model review and test every branch before it reaches you.
Open-source maintainers: use Pro free on a registered public repository.
Thrum User Guides
-
Install Thrum for free; no account is needed.
-
Start a coordinator and, when you ask, a reviewer on another model.
-
Describe your request to the coordinator and let it hand the work to your agents.
-
Read its report, and choose whether it asks before each merge or merges routine changes itself.
Thrum Pricing
Free with no account. Pro: $15/month or $120/year, with a 14-day trial and no card; buying opens soon. Pro is free for each registered public repository.
Thrum launch and community
Nov 9, 2026 on Hacker News · 6 points, 7 comments
“I built Thrum which runs the coding agents you already use and gives them functionality to work as a team.”
What people asked
How does Thrum keep 100 or more agents productive?
The maker said he uses a hierarchy: a main coordinator makes the major decisions, other coordinators manage agents on their own virtual machines, and a steward watches specialty agents and keeps him updated.
Thrum limitations
- Buying Pro was not open yet at launch; only the 14-day trial was available.
- It is built for people running more than one coding agent; the site itself says one agent with helpers may not need it.
- SSH connections are not supported yet; the maker says they are coming.
Thrum Frequently Asked Questions
Is Thrum free?
Which agents does it work with?
Does my code go to a Thrum cloud?
How does it handle many agents?
Thrum alternatives
How Thrum compares
- Timon: Timon runs several agent CLIs in one desktop window with a Kanban board; Thrum lets agents message each other under one coordinator.
Timon
A free desktop app that runs Claude Code, Codex, Grok, Copilot and other coding agents side by side, with a Kanban pipeline that codes, reviews and merges.
AI Code Assistant · Free
SEO.party
A free library of community SEO skills for coding agents such as Claude Code, Codex and Cursor, all built on OpenSEO's data tools.
AI SEO Assistant · Free
Is Software Solved?
A free public index that scores how close AI is to doing software work on its own, from 28 signals such as jobs, AI usage, revenue, security and outages.
Research Tool · Free
Arc UI
A library of 239 animated React components and 138 blocks for shadcn and Next.js, added with the shadcn CLI or by an AI agent through its MCP server.
AI Developer Tools · Freemium, from $149/year
Foundry AI
Foundry AI lets businesses build, evaluate, and improve AI agents for tasks like customer support and sales, with real-time feedback and continuous improvement for reliable outcomes.
AI Developer Tools
Nfig
Nfig enables AI agents to browse, click, and perform tasks on any website using natural language commands, simplifying web automation and unlocking new use cases with its easy-to-integrate API.
No-Code&Low-Code