Browser-based graphical shell and control plane for a Linux server you own — real terminal, file manager, live metrics and BYOK AI in one mobile-first workspace. Next.js 16 · React 19 · node-pty · Tailwind v4 · shadcn/ui. Public alpha, self-hosted.
TypeScript browser-shellcontrol-planefile-managerhomelabmobile-firstnextjsptyreactself-hostedserver-managementshadcn-uisysadmintailwindcssterminaltypescriptvpsweb-desktopweb-terminal
import { MSO_BLOCK_URI } from "./ui-block";
import { type McpTool, READ_ONLY, S, str } from "./tool-kit";
const SCALAR_SCHEMA = {
anyOf: [{ type: "string" }, { type: "number" }, { type: "boolean" }, { type: "null" }],
} as const;
const FIELD_SCHEMA = {
type: "object",
properties: {
id: { type: "string", minLength: 1, maxLength: 80 },
label: { type: "string", minLength: 1, maxLength: 160 },
value: SCALAR_SCHEMA,
input: { type: "string", enum: ["text", "number", "email", "url", "textarea", "boolean"] },
editable: { type: "boolean" },
required: { type: "boolean" },
placeholder: { type: "string", maxLength: 240 },
},
required: ["id", "label"],
additionalProperties: false,
} as const;
const CHECK_SCHEMA = {
type: "object",
properties: {
label: { type: "string", minLength: 1, maxLength: 200 },
state: { type: "string", enum: ["pass", "warn", "fail", "info"] },
detail: { type: "string", maxLength: 500 },
},
required: ["label", "state"],
additionalProperties: false,
} as const;
const OUTPUT_SCHEMA = {
type: "object",
properties: {
label: { type: "string", minLength: 1, maxLength: 160 },
value: SCALAR_SCHEMA,
},
required: ["label", "value"],
additionalProperties: false,
} as const;
const ACTION_SCHEMA = {
type: "object",
properties: {
id: { type: "string", minLength: 1, maxLength: 80 },
label: { type: "string", minLength: 1, maxLength: 120 },
style: { type: "string", enum: ["primary", "secondary", "danger"] },
prompt: {
type: "string",
minLength: 1,
maxLength: 1200,
description: "Follow-up instruction sent only after the user presses this button. Normal tool scope, approval, and audit rules still apply.",
},
confirm: { type: "string", maxLength: 240 },
},
required: ["id", "label", "prompt"],
additionalProperties: false,
} as const;
export const BLOCK_OUTPUT_SCHEMA = {
// Output is normalized by normalizeBlock() below. Keep the advertised result
// envelope intentionally compact: duplicating every input constraint here adds
// several KiB to tools/list without adding a runtime safety boundary.
type: "object",
properties: {
kind: { type: "string" }, title: { type: "string" }, description: { type: "string" }, status: { type: "string" },
fields: { type: "array" }, checks: { type: "array" }, outputs: { type: "array" }, actions: { type: "array" },
},
required: ["kind", "title", "status", "fields", "checks", "outputs", "actions"],
additionalProperties: false,
} as const;
const text = (value: unknown, max: number, fallback = "") => {
if (typeof value !== "string") return fallback;
const clean = value.trim();
return clean ? clean.slice(0, max) : fallback;
};
const scalar = (value: unknown): string | number | boolean | null => {
if (typeof value === "string") return value.slice(0, 2_000);
if (typeof value === "number" && Number.isFinite(value)) return value;
if (typeof value === "boolean" || value === null) return value;
return "";
};
const choice = <T extends string>(value: unknown, allowed: readonly T[], fallback: T): T =>
typeof value === "string" && allowed.includes(value as T) ? value as T : fallback;
function rows(value: unknown, max: number): Array<Record<string, unknown>> {
return Array.isArray(value)
? value.filter((item): item is Record<string, unknown> => Boolean(item) && typeof item === "object" && !Array.isArray(item)).slice(0, max)
: [];
}
function normalizeBlock(input: Record<string, unknown>) {
const fields = rows(input.fields, 20).map((field, index) => ({
id: text(field.id, 80, `field-${index + 1}`),
label: text(field.label, 160, `Field ${index + 1}`),
value: scalar(field.value),
input: choice(field.input, ["text", "number", "email", "url", "textarea", "boolean"] as const, "text"),
editable: field.editable !== false,
required: field.required === true,
...(text(field.placeholder, 240) ? { placeholder: text(field.placeholder, 240) } : {}),
}));
const checks = rows(input.checks, 20).map((check) => ({
label: text(check.label, 200, "Check"),
state: choice(check.state, ["pass", "warn", "fail", "info"] as const, "info"),
...(text(check.detail, 500) ? { detail: text(check.detail, 500) } : {}),
}));
const outputs = rows(input.outputs, 20).map((output) => ({
label: text(output.label, 160, "Result"),
value: scalar(output.value),
}));
const actions = rows(input.actions, 8).map((action, index) => ({
id: text(action.id, 80, `action-${index + 1}`),
label: text(action.label, 120, "Continue"),
style: choice(action.style, ["primary", "secondary", "danger"] as const, "secondary"),
prompt: text(action.prompt, 1_200, "Continue with the selected MSO action."),
...(text(action.confirm, 240) ? { confirm: text(action.confirm, 240) } : {}),
}));
return {
kind: choice(input.kind, ["validation", "crud", "action"] as const, "action"),
title: text(input.title, 240, "MSO block"),
...(text(input.description, 1_000) ? { description: text(input.description, 1_000) } : {}),
status: choice(input.status, ["neutral", "pending", "success", "warning", "error"] as const, "neutral"),
fields,
checks,
outputs,
actions,
};
}
export const BLOCK_TOOLS: McpTool[] = [
{
name: "render_mso_block",
title: "Render MSO Block",
description:
"Render a compact MCP App block only when a visible interaction helps: validation state, action buttons, or CRUD input-output. " +
"workflow_start is headless and never opens this automatically. Buttons send a user-approved follow-up to ChatGPT; the actual read/write/exec operation still uses the normal scoped tool, approval, and audit path. Raw HTML and arbitrary executable actions are not accepted.",
chatgptDescription:
"Render a compact validation/action/CRUD block. Use it only for user-facing confirmation or structured input-output; workflow_start stays invisible. Button presses return a follow-up to the chat and do not bypass tool approvals.",
scope: "read",
annotations: READ_ONLY,
inputSchema: S({
kind: { type: "string", enum: ["validation", "crud", "action"] },
title: { type: "string", minLength: 1, maxLength: 240 },
description: { type: "string", maxLength: 1_000 },
status: { type: "string", enum: ["neutral", "pending", "success", "warning", "error"] },
fields: { type: "array", maxItems: 20, items: FIELD_SCHEMA },
checks: { type: "array", maxItems: 20, items: CHECK_SCHEMA },
outputs: { type: "array", maxItems: 20, items: OUTPUT_SCHEMA },
actions: { type: "array", maxItems: 8, items: ACTION_SCHEMA },
}, ["kind", "title"]),
outputSchema: BLOCK_OUTPUT_SCHEMA,
meta: {
ui: { resourceUri: MSO_BLOCK_URI, visibility: ["model", "app"] },
"ui/resourceUri": MSO_BLOCK_URI,
"openai/outputTemplate": MSO_BLOCK_URI,
"openai/toolInvocation/invoking": "Preparing interaction…",
"openai/toolInvocation/invoked": "Interaction ready",
"openai/widgetAccessible": true,
},
run: async (input) => normalizeBlock({ ...input, title: str(input, "title") }),
},
];