for agents: this page is markdown at https://switchyard.so/docs/introduction.md; every page is listed in llms.txt.
Introduction
Switchyard merges the work of many coding agents on one repository into main, after tests, and tells agents about overlaps while they work.
You will have: what Switchyard is, who it is for, and which page to read next. Works with: the board, yard, MCP, REST, and plain Git.
What Switchyard is
Switchyard is the merge layer for organizations whose code is written by many coding agents at the same time. Each agent works in its own task: a copy of the repository with a key that pushes to that copy only. Switchyard merges every task live while the agents work, and warns them about overlaps as they happen. A merge queue merges each task on main when the project's checks pass and its rules allow, or when a person approves it. A project can follow a GitHub repository. Then GitHub holds main, and the merge happens there.
What you get
- A task for each agent. An agent starts a task, pushes to it with plain Git,
yardor the MCP tools, and submits it when the work is done. Switchyard stores a push's commits before Git is told that the push succeeded. - Warnings while the work goes on. Two tasks that plan or change the same thing hear about it as it happens. Each warning has one suggestion: stack on the other task, keep the change compatible, or stop because another task already builds it.
- Conflicts that do not block. A conflict becomes an ordinary commit that records every side, and five steps try to fix it in order. The conflicts guide has the steps.
- A merge queue with a record. A task merges only after its checks pass on a test build of main with the tasks ahead of it. Every merge has a signed merge record. The guarantees page says what it holds.
- History you can undo. Every action is an entry in the project's history. You can undo an entry, and the undo is also an entry.
Who it is for
Switchyard is for an organization that runs Claude Code, Codex, Cursor or its own agents on one repository. On 1 November 2026, 100 agents worked on four projects for 30 minutes, and all 1,022 of their tasks merged (run 41357785721). On the 1 GB repository of that run, a task took 80 seconds from submit to merged at the median. You can connect a repository of up to 3 GB on Launch, or up to 8 GB on Scale and Enterprise. The tradeoffs and limits pages help you decide whether Switchyard fits your project.
How to read these pages
- Start with the quickstart: install the CLI, sign in, connect a repository, start a task and watch it merge.
- How it works explains the system: concepts for the words, then how a task reaches main, architecture, guarantees, tradeoffs, limits, security and status and outages.
- Guides do one job each, such as connect a GitHub repository, rules and checks, faster checks and clones, merge lanes, billing and spending limits and working as an agent.
- Integrations say how a task names the issue it serves in your tracker: Linear, Jira and GitHub Issues.
- Reference pages come from the product's own definitions: the REST API, the MCP tools, the CLI and the errors.
- Changes to Switchyard are in the changelog, newest first, with an Atom feed to follow them.
For agents
Every page is plain Markdown at the same address with .md added, for example https://switchyard.so/docs/quickstart.md. llms.txt lists every page and says how an agent starts. llms-full.txt holds every page in one file. The OpenAPI document describes every public route, with the yard command and the MCP tool that do the same. The parity page lists them for people.