A programmable Windows browser for AI-assisted web automation
Script WebView2 browser workflows, connect MCP-capable AI tools, and run headed remote executors through Remote Ops. Browser actions stay inspectable, tool calls are logged, and secrets never leave your machine.
Watch a two-screen desktop workflow where Codex uses Guida to inspect rendered career sites, compare page behavior, and capture findings. Guida is useful when a coding assistant needs a visible, approval-gated browser alongside Playwright or Puppeteer checks.
2 min 56 secTwo-screen captureLocal workspace notes
Script pages, connect AI tools, process queues, and keep sensitive actions inspectable.
Programmable Browser Automation
Drive WebView2 browser tabs with scripts: navigate pages, click elements, fill forms, extract content, and capture screenshots.
AI and MCP Control
Connect MCP-capable tools so AI agents can inspect tabs, request browser actions, and run scripts with your oversight.
Complete, Node, and Remote Ops
Use Guida Complete for the full operator UI, Guida Node for headed remote execution, and Remote Ops as the JSON-RPC control plane. Remote Ops also lays the groundwork for future browser or desktop control surfaces.
Built-in Script Engine
JavaScript, Lua, and Janet with 160+ API methods, IntelliSense, debugging, and workspace-aware execution.
Workspaces and Data Pipelines
Persist data in workspace storage, queues, full-text search, PostgreSQL, MongoDB, S3-compatible storage, and Elasticsearch.
Workspace Backups
Create full workspace archives or smaller managed-state backups with consistent SQLite snapshots for queues and workflow ledger history.
Approval and Domain Controls
Review AI-requested actions before they run, approve once or for the session, and restrict navigation to allowed domains.
Audit Trail and Secret Isolation
MCP actions are logged, secrets are encrypted with DPAPI, and MCP-origin scripts cannot read decrypted secret values.
Recipes from real scripts
Start from workflows that ship with Guida, then adapt them to your own workspace.
Point any MCP-compatible AI at Guida's server. Or use Ollama and LlamaSharp directly inside the app.
2
Stay in control
See exactly what the AI wants to do before it runs. Approve once, trust for the session, or let it fly - your call.
3
Automate with confidence
Reviewed scripts run unattended. Workspaces persist data. Queue workers process items in parallel.
The g.* scripting API
Navigate, extract, and store data with a clean scripting API. Write scripts yourself, or let the AI generate them.
scrape-products.js
async functionmain() {
try {
await g.nav.navigateToUrl("https://example.com/products");
await g.dom.waitForDomStable();
// Extract name and price from every product cardconst products = await g.dom.getElements(
".product", ["innerText", "data-price"]
);
// Store each product in the workspacefor (const p of products) {
g.store.put("products", p.innerText, p);
}
g.log("Stored " + products.length + " products");
} catch (e) {
g.log("Error: " + e.message);
}
}
g.nav.navigateToUrl("https://example.com/products")
g.dom.waitForDomStable()
-- Extract name and price from every product cardlocal products, err = g.dom.getElements(
".product", "innerText,data-price"
)
if err then
g.log("Error: " .. err)
returnend-- Store each product in the workspacefor i, p inipairs(products) do
g.store.put("products", p.innerText, p)
end
g.log("Stored " .. #products .. " products")
(g.nav.navigateToUrl"https://example.com/products")
(g.dom.waitForDomStable)
# Extract name and price from every product card
(def products (g.dom.getElements".product""innerText,data-price"))
# Store each product in the workspace
(each p products
(g.store.put"products" (get p "innerText") p))
(g.log (string"Stored " (length products) " products"))
Ready to take control?
Download Guida, connect your AI, and start automating - with full visibility into every action.
