--- title: "Cork vs qm: a team agent harness, or a coordination layer for coding agents · Cork" description: "Cork vs qm: YC's open source qm is a multiplayer agent harness for Slack and web. Cork keeps parallel AI coding agents from overwriting each other's code." canonical: "https://corkcli.com/blog/cork-vs-qm" last_updated: "2026-11-14" published: "2026-10-03" author: "Kevin Vera" --- # Cork vs qm: a team agent harness, or a coordination layer for coding agents Cork vs qm: YC's open source qm is a multiplayer agent harness for Slack and web. Cork keeps parallel AI coding agents from overwriting each other's code. By Kevin Vera · October 3, 2026 · [Comparisons](https://corkcli.com/blog/c/comparisons) · 7 min read How is Cork different from qm? qm, open sourced by Y Combinator, is a multiplayer agent harness for work: a company deploys it in its own cloud, and people use agents from Slack and the web for all kinds of work, including engineering. Cork is narrower and sits at a different layer: it keeps several AI coding agents, and the people running them, from overwriting each other on one codebase, and briefs each agent with what the others already learned. qm is where a team's agents live and work; Cork is what keeps coding agents from colliding when they edit the same repository. They solve different problems, and a team could reasonably run both. ## Cork vs qm at a glance | Topic | Cork | qm (as of October 2026) | |---|---|---| | What it is | [Coordination layer](https://corkcli.com/blog/coordination-layer-ai-coding-agents) for parallel AI coding agents and teammates | Multiplayer agent harness for work, in Slack and on the web | | Where agents run | Where they already run, on your machine | In per-scope sandboxes in your own cloud (Fly or AWS) | | Agents covered | Claude Code, Codex, Cursor (hooks); any MCP client | Pi, OpenCode, Codex and Claude Code drive the same core | | Scope of work | Code: shared repositories | General work: Slack, web apps, crons, webhooks, repositories | | What is shared | Signatures, callers, recent changes, notes, live edits between agents | Shared channels, projects and skills; scoped memory per person and room | | Conflict handling | Write gate, live change notices, [semantic lint](https://corkcli.com/blog/semantic-lint), rebase-and-test landing | Isolated sandboxes per scope; we found no code-level conflict coordination | | Install | `curl -fsSL https://corkcli.com/install.sh \| sh` | Deploy with the qm CLI to your own cloud account | | Open source | See corkcli.com | Yes, MIT | | Pricing | Free (local, and 10 agent hours a month in the cloud); $20 a month per 100 agent hours, pooled | Free software; we could not verify published pricing for a hosted plan | ## What is qm? qm describes itself as "a multiplayer agent harness for work. In Slack and on the web. Run it in your own cloud, with your own models and keys." It is published by YC at github.com/yc-software/qm under the MIT license. Each person and each room gets its own scoped memory, files, keychain view, permissions, crons, web apps and a durable sandbox, described as "its durable computer, where installed tools stay installed." People keep a consistent identity between Slack and the web, can share skills, and can run background work through crons, watches and inbound webhooks. The agent "acts as the person it's working for, with their credentials and permissions, and everything it does is audited." qm is not tied to one model vendor: "Pi, OpenCode, Codex, and Claude Code all drive the same core." For engineering, its README lists working in an existing repository: running tests, opening PRs, monitoring CI and checking logs. The core is TypeScript on Node with Fastify, Postgres for state and Bolt for Slack, deployed to Fly or AWS through its CLI. ## What is Cork? Cork (corkcli.com, by Lithometric) keeps several AI coding agents, and the people running them, from overwriting each other on one codebase, and hands each agent what the others already learned. It does not run agents or host machines: Claude Code, Codex and Cursor keep running where they already run, and Cork coordinates them. Hooks in Claude Code, Codex and Cursor run the protocol with no extra model calls, and an MCP server (https://mcp.corkcli.com) serves any MCP client. Agents get a briefing with the exact signatures, callers and recent changes of the code they are about to touch. When another agent renames a function yours depends on, your agent hears it on its next step. A [write gate](https://corkcli.com/blog/write-gate) holds an edit against code another agent is changing until your agent has their version. Semantic lint turns a teammate's rename into a rule, landing turns `git push` into rebase, retest and push, and agents can message each other or whoever is working on a given function. Cloud plans add a live code graph and a team dashboard. ## Where qm stands out - **Agents for the whole company.** qm is built for work in general, not only code, with web apps, shared skills and background jobs, and it meets people in Slack where they already talk. - **Your cloud, your keys.** Every deployment runs in the operator's own cloud account with its own models and keys, and it is MIT licensed. - **Durable, scoped sandboxes.** Each person and room gets a persistent computer, memory and permissions, which suits long-running and background work. - **Harness choice.** Pi, OpenCode, Codex and Claude Code all drive the same core, so you can switch without rebuilding the scaffolding. - **Auditable identity.** Agents act with the requesting person's credentials, and their actions are audited. ## Where Cork wins - **Code-level coordination.** qm isolates scopes in sandboxes; it does not, as far as we could find, tell one coding agent that another just changed a signature it uses. Cork does that live, and gates writes to code another agent is changing. - **Fewer agent merge conflicts.** Semantic lint and rebase-and-test landing catch the rename that would otherwise break a teammate's branch. - **Agents arrive informed.** Cork's briefings hand agents the signatures and callers they need. In Cork's benchmark (12 Claude Code agents at once on one repository, hard tasks), agents with Cork used 53% fewer tokens than without it. - **Nothing to host.** Cork installs with one command and works with agents where they already run; qm is a deployment you operate. ## Can you use both? Yes, for different jobs. qm can be a team's general agent harness in Slack and the web, while Cork coordinates the coding agents developers run on their machines. Because qm can drive Claude Code and Codex, it may also be possible to point those agents at Cork's MCP server, which any MCP client can use; we have not verified that setup, so treat it as something to test. If your team only needs Slack-based assistants and never has several agents editing one repository at once, Cork adds little. ## Pricing compared - **qm:** free, MIT-licensed software you deploy to your own Fly or AWS account; you pay your own cloud and model costs. We could not verify published pricing for any hosted qm offering (as of October 2026). - **Cork:** Cork Free gives unlimited agents and repos on one machine with no account and all data on the machine. Cloud Free adds an account with 10 agent hours a month (30 bonus hours the first month) and a 30-day history window. Beyond that you pay for agent hours, not seats: $20 a month per unit of 100 hours, pooled across the workspace, graduated to $17 (units 11 to 50) and $15 (51 and up), with 15% off a year. Every member is a player, with teammates' machines, prompts matched to code by meaning, shared knowledge and 90-day history. Enterprise starts from about 100 units. ## Bottom line qm is a broad, open source harness that gives a whole company multiplayer agents in Slack and the web, running in its own cloud. Cork is a focused coordination layer for multi-agent coding: it keeps parallel coding agents and teammates from overwriting each other and briefs them on the code they touch. If your problem is "give everyone agents," look at qm. If your problem is "our coding agents keep stepping on each other," that is what Cork is for. ## Frequently asked questions ### Is Cork an alternative to qm? Only partly. qm is a general agent harness for team work; Cork is a coordination layer for coding agents. They overlap only in that both are multiplayer, and many teams could use both. ### Does qm stop coding agents from overwriting each other? qm gives each person and room an isolated sandbox. We did not find a feature that coordinates live edits between agents working on the same repository, which is the gap Cork fills. ### Is qm open source? Yes. qm is published by YC at github.com/yc-software/qm under the MIT license. ### Which agents does qm support? Its README says Pi, OpenCode, Codex and Claude Code all drive the same core. ### Do I need to host anything to use Cork? No. Cork installs with one command and works with Claude Code, Codex and Cursor where they already run. Cork Free keeps all data on your machine. ## Sources Checked October 2026: - - - ## Related reading - [Cork vs Amika: VMs for your agents, or coordination between them](https://corkcli.com/blog/cork-vs-amika): Cork vs Amika: Amika provisions multiplayer VM Rigs for coding agents. Cork keeps parallel agents on one codebase from overwriting each other. - [Cork vs HumanLayer: an agent IDE for teams, or coordination under any tool](https://corkcli.com/blog/cork-vs-humanlayer): Cork vs HumanLayer: HumanLayer is a multiplayer coding agent IDE and cloud; Cork keeps parallel agents from overwriting each other wherever they run. - [Cork vs Mosaic: shared session memory, or live coordination on the code](https://corkcli.com/blog/cork-vs-mosaic): Cork vs Mosaic: Mosaic syncs AI coding sessions into shared team memory; Cork keeps parallel agents from overwriting each other. How they differ and fit. - [Cork vs Superconductor: a hosted agent workspace, or coordination for the agents you run](https://corkcli.com/blog/cork-vs-superconductor): Cork vs Superconductor: Superconductor runs coding agents in cloud sandboxes for teams. Cork stops parallel agents from overwriting each other's code. [All articles](https://corkcli.com/blog.md) · [Cork in Markdown](https://corkcli.com/index.md) · [llms.txt](https://corkcli.com/llms.txt) This page in HTML: https://corkcli.com/blog/cork-vs-qm