Explainer library
These pages explain the ideas and technology behind Ronin. Each one answers a real question on its own; together they map the machine.
What Ronin does to your Agent: nothing inside, everything at the edges. Coworking is coordination between Agents, not capability added to one. Three ways Ronin shapes an Agent, what coworking is and is not, and what an ack looks like.
Start here →A Workbench scene with two Agent providers, two independent workspaces, and one locally run coworkspace around them.
See the Workbench →Agents can work independently or use optional messages, shared surfaces, and Routines when a job benefits from coordination.
See the coordination loop →Start with a bare Agent, then compare the Cowork floor, Ronin Base, managed file coordination, and additional installed capabilities.
Understand the map →Keep your providers, Agent CLIs, repositories, ordinary files, and development workflow outside any proprietary Ronin format.
See the ownership and data boundary →Ronin offers convenient tools at the edges of cooperation without becoming the Agent's prompt engine or prescribed methodology.
See coordination without control →Select and customize reading, SOPs, Macros, Actions, tools, and connections in owner stores that Ronin updates do not overwrite.
See how owner shadows work →The browser can close; the agent keeps working. How Ronin uses a tmux server to keep live agent terminals running on your machine and reconnect the same session later.
Read the explainer →Put the Ronin machine, your computer, and your phone on the same private network, and open your coworkspace from anywhere without exposing it to the public internet.
Read the guide →Choose a cloud provider and the right CPU, memory, storage, Linux image, SSH, and private-network setup for an always-on team of AI coding agents. No GPU or dedicated AI server required.
Read the explainer →Why several Agents collide in one checkout and how a worktree per Agent fixes it, chosen once per repository. Then the whole journey of using it in Ronin: mark a lead, commit and hand in, the lead's promotion, and your merge of the release pull request.
Read the explainer →Benchmarks compare capability. This opening proposal maps five observable working personalities—and asks how longer work episodes might test what an Agent is actually like to collaborate with.
Read the discussion →Run your Ronin coworkspace on the machine that will host the sessions.
