Separate task modes
Study the division of agent responsibilities when deciding which workflow stages should inspect, plan, or change a repository.
Study OpenCode’s coding-agent workflow to improve how your own tasks move from investigation to changes and review. Use its interaction and extension design as references for development work organized in Floatboat.
OpenCode is an open-source AI coding agent published by Anomaly. It connects task conversations with repository tools. The model service supplies reasoning and generation, the execution environment handles files and commands, and the application coordinates the workflow.
For workflow design, OpenCode is a concrete implementation to study rather than a requirement to switch products. Its agent modes and interactive feedback can inform how you separate investigation from implementation, expose a change for review, and add task-specific integrations to your own environment.
The project source is MIT licensed. Model-provider accounts, subscription access, remote services, and community additions have separate requirements. The source can be inspected and adapted without implying that all associated hosted services are free or included.
Understand what the project demonstrates, then decide which mechanism addresses a real gap in your own workflow.
Study the division of agent responsibilities when deciding which workflow stages should inspect, plan, or change a repository.
Use the interactive workflow as a reference for showing progress and tool results where a person can make an informed decision.
Inspect an integration’s tools and maintainer before adapting the pattern to the development workflow you actually need.
Use these scenarios as design references for work organized in Floatboat. Any software integration needs its own compatibility and permission checks.
Define what an agent must learn before implementation begins, including relevant files, constraints, and unresolved decisions.
Separate the proposed change from acceptance so generated code is paired with an inspectable diff and relevant checks.
Study the extension interfaces when considering a connection to a project’s existing checks or external development systems.
Start with a task and a small testable change, whether that means better instructions, a new tool connection, or a runtime component.
Write down what investigation must deliver before planning or implementation can safely continue.
Match tool access and the active mode to each stage rather than giving every step the same ability to change files.
Evaluate whether the workflow makes errors visible and gives the user enough evidence to accept or revise the change.
Compare independent agents through their interfaces, extension models, and approach to carrying work forward.