{"id":23895,"date":"2026-09-11T00:56:13","date_gmt":"2026-09-11T00:56:13","guid":{"rendered":"https:\/\/scannn.com\/vastsa-pi-desktop-local-first-ai-coding-agent-desktop-electron-rust-host-core-pi-agent-harness-user-installable-plugins-%c2%b7-github\/"},"modified":"2026-09-11T00:56:13","modified_gmt":"2026-09-11T00:56:13","slug":"vastsa-pi-desktop-local-first-ai-coding-agent-desktop-electron-rust-host-core-pi-agent-harness-user-installable-plugins-%c2%b7-github","status":"publish","type":"post","link":"https:\/\/scannn.com\/lv\/vastsa-pi-desktop-local-first-ai-coding-agent-desktop-electron-rust-host-core-pi-agent-harness-user-installable-plugins-%c2%b7-github\/","title":{"rendered":"vastsa\/PI-Desktop: Local-first AI coding agent desktop: Electron + Rust host core + pi Agent Harness + user-installable plugins \u00b7 GitHub"},"content":{"rendered":"\n
\n\n
\n

<\/svg>Important<\/p>\n

PI-Desktop is currently in Early Preview.<\/strong><\/p>\n

The project is actively developed and already usable for real coding workflows, but APIs, extension interfaces, and some desktop behaviors may continue to evolve.<\/p>\n<\/div>\n

Most coding agents live inside a terminal, an editor extension, or a hosted service.<\/p>\n

PI-Desktop gives them a workspace of their own.<\/p>\n

<\/p>\n\n\n\n\n
\n

Work across repositories and sessions without tying your agent workflow to one editor or terminal.<\/p>\n

Projects, conversations, reviews, files, previews, notifications, and extensions live in one workspace.<\/p>\n<\/td>\n

\n

Use OpenAI, Anthropic, local models, hosted gateways, or any OpenAI-compatible API.<\/p>\n

Configure multiple providers and models, then switch between them per session.<\/p>\n<\/td>\n<\/tr>\n

\n
\n

Inspectable by default<\/h3>\n

<\/svg><\/a><\/div>\n

Agents can read files, edit code, and run commands \u2014 but privileged actions pass through PI-Desktop’s permission layer.<\/p>\n

Review diffs, inspect command output, and decide how much autonomy each session gets.<\/p>\n<\/td>\n

\n

Add Skills, MCP servers, Subagents, and installable Plugins.<\/p>\n

Plugins can contribute tools, commands, panels, themes, services, skills, and new workspace experiences.<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

<\/markdown-accessiblity-table><\/p>\n\n

Getting started only takes a few steps:<\/p>\n

    \n
  1. \n

    Connect a model<\/strong>
    \nOpen Settings \u2192 Model configuration<\/strong>, choose a provider or compatible API, and add your credentials.<\/p>\n<\/li>\n

  2. \n

    Open a project<\/strong>
    \nAdd any local repository or project directory from the sidebar.<\/p>\n<\/li>\n

  3. \n

    Pick Agent, Plan, or Goal<\/strong>
    \nAgent<\/strong> starts working. Plan<\/strong> waits until you approve a frozen implementation plan. Goal<\/strong> waits until you approve the outcome, then the agent chooses the path.<\/p>\n<\/li>\n

  4. \n

    Review the result<\/strong>
    \nInspect edits in the Review panel, check command output, preview the application, and continue the conversation without leaving PI-Desktop.<\/p>\n<\/li>\n<\/ol>\n\n

    Same agent. Three gates. Privileged tools still go through the permission layer in every mode.<\/p>\n

    <\/p>\n\n\n\n\n\n\n\n\n
    Agent<\/strong><\/th>\nPlan<\/strong><\/th>\nGoal<\/strong><\/th>\n<\/tr>\n<\/thead>\n
    You approve<\/td>\nNothing extra<\/td>\nThe implementation plan<\/td>\nThe outcome and acceptance criteria<\/td>\n<\/tr>\n
    The agent does<\/td>\nReads, edits, runs commands, tests, iterates<\/td>\nStudies the repo, writes a frozen plan, then waits<\/td>\nPicks the path and works until the goal is met<\/td>\n<\/tr>\n
    Use when<\/td>\nYou want it to just do the work<\/td>\nThe change is large or risky and you want the approach first<\/td>\nYou care about the result, not the route<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    <\/markdown-accessiblity-table><\/p>\n

    Agent<\/strong> is the default loop: inspect the tree, patch files, run commands, and keep going.<\/p>\n

    Plan<\/strong> is the approval boundary. The agent researches first and produces an immutable implementation plan. Execution does not start until you sign off.<\/p>\n

    Goal<\/strong> is outcome-first. You lock the objective and acceptance criteria; the agent decides how to get there.<\/p>\n\n

    \n

    Delegate work to Subagents<\/h3>\n

    <\/svg><\/a><\/div>\n

    Large tasks rarely belong in one context window.<\/p>\n

    PI-Desktop can delegate independent work to background Subagents for things like:<\/p>\n