EverDock Desktop

EverDock is built around the idea that AI coding agents are becoming software workers, not just chat interfaces. A single agent in a single terminal can be managed manually, but multiple agents working across repositories and servers need orchestration, visibility, approval enforcement, durable sessions, event replay, and cost controls. The product is currently being tested inside real developer workflows to improve how engineers assign tasks, monitor long-running agent sessions, review diffs and test evidence, approve risky actions, and continue work from different devices.

Modern coding agents can read codebases, edit files, run shell commands, install dependencies, create diffs, run tests, and deploy software. That power creates a new operations problem: developers need to know which agent is doing what, where it is running, what files changed, which commands were executed, whether a session is stuck, how much compute or token budget is being consumed, and which actions require human approval. Terminal tools, SSH clients, and remote desktops solve access, but they do not provide a control plane for AI workers.

Founder / Product Builder / Desktop Application Engineer

EverDock uses a Tauri/React desktop workstation with a Rust core, local state, terminal/process abstractions, and a planned Runner layer that executes work on local machines, SSH servers, or managed servers. The product model centers on Account, Space, Workspace, ExecutionTarget, WorkspaceBinding, Runner, Session, Pane, TaskRun, AgentSession, ActionIntent, Approval, Policy, Artifact, FileChange, Diff, and Event. Risky actions are meant to be enforced at the Runner executor rather than only in the UI, so approvals, budgets, command policy, secret access, and audit trails remain part of the execution boundary.