Chatbots answer.
Appilots runs your app.
Appilots connects an AI agent to your React Native UI through a generated MCP map. It navigates, fills forms, and confirms actions — completing real tasks, not just talking about them.
Your users ask.
Your app does the rest.
Appilots reads a generated MCP map of your React Native app and drives the real interface — navigate, fill, press, confirm. Embeddable in minutes, themed to your product.
An agent in your
app, in 3 lines.
Generate the MCP document, wrap your navigation, and ship. Appilots's agent reads your app's map and drives the UI — no per-screen wiring, no glue code.
Not another sidebar chatbot.
Most "AI assistants" can only talk about your app. Appilots operates it — the same screens, forms, and buttons your users already tap.
Tells the user what to do
- Answers in a separate window, then leaves the work to the user
- Has no idea what screen the user is on or what's possible next
- Needs custom tools and glue code wired for every action
- Copy-paste steps, dead ends, and "I can't do that here"
Does it, inside your app
- Navigates, fills, and presses in the real UI to finish the task
- Reads the MCP map, so it knows every screen, form, and action
- One provider, zero per-screen wiring — the map is generated for you
- Pauses for an explicit confirm before anything sensitive
Three steps from install to a working agent.
Appilots doesn't bolt a chatbot onto your app. It builds a structured map of your screens and lets the agent operate the real UI.
Generate the MCP document
The CLI walks your navigation graph, screens, forms, and actions and emits a Model Context Protocol map of the app.
$ npx appilots init
Embed the SDK
Wrap your navigation with one provider. The chat surface themes itself to your app — colors, radii, dark mode — out of the box.
<AppilotsProvider />
The agent drives the UI
Users ask in plain language. The agent reads the map, plans the steps, and operates the real interface — pausing to confirm anything sensitive.
navigate · fill · press · confirm
Everything you need to run an agent in production.
A developer dashboard, an embeddable SDK, and the guardrails to ship an in-app agent your users can trust.
MCP map of your app
An auto-generated, structured model of every screen, route, and form — the context the agent reasons over.
Drives the real UI
Not a sidebar chatbot. The agent navigates, fills required inputs, presses buttons, and completes flows in the live interface.
Embeddable chat surface
A React Native bottom-sheet chat that themes itself to the host app. One FAB, light and dark tokens, zero design work.
Permissions & confirmations
Mark actions as sensitive and the agent pauses for an explicit approve / reject before it touches anything destructive.
Budget & token controls
Set monthly spend caps per project, watch token usage in real time, and get alerts before you hit the ceiling.
Insights & logs
Every run is a readable action log. See what the agent did, where it paused, and which prompts drove the best outcomes.
An assistant that completes the task — not just answers about it.
Drop in the SDK and users get an in-app agent that turns a sentence into a sequence of real actions, with a live breadcrumb of every step.
- Action breadcrumb shows navigate, fill, and press steps as they run — like a build log for the UI.
- Confirm before destructive actions. The agent drafts the change and waits for the user to approve.
- Themed to your app with light and dark tokens. Swap the assistant avatar and you're done.
A fast, quiet control room for your agents.
Manage projects, AI providers, API keys, the MCP map, prompts, and budgets from a keyboard-friendly dashboard — no decorative chrome, just the work.
- Projects & providers — point a project at OpenAI or Anthropic and rotate keys without a redeploy.
- MCP map, prompts & sandbox — inspect the generated app model and test runs before they ship.
- Budgets & insights — spend caps, token usage, and a readable history of every agent run.
Bring your own model.
Appilots is provider-agnostic. Connect the model you already pay for — your keys, your data, your spend.
baseURL and your key in the dashboard. Your keys, your data, your spend.
Two commands and one provider.
Install the SDK, generate your app's MCP document, and wrap your navigation. The agent picks up the map automatically.
Questions, answered.
Is this just a chatbot in my app?
Which AI models can I use?
Where does my data go?
How much work is the integration?
npx appilots init to generate the map, and wrap your navigation with one provider. No per-screen wiring — the agent picks up the map automatically.What does early access include?
Get early access to Appilots.
Tell us a bit about you and we'll review your request for private beta access.