# Agents, skills and MCP servers | Runner AI > Runner's agent builds and runs your store. Give it your methods as skills, connect the apps you already use, and point it at your own MCP servers. Canonical: https://www.runnerai.com/agents ## One agent. Your way of working. Runner's agent plans the work, hands each part to a built-in specialist, and uses the skills, apps and servers you give it. You decide what it does alone, and what it asks first. ### 1. Ask in plain words Tell the agent what you need in chat. It plans the work and hands each part to the right built-in agent: Store Designer, Data Analyst, SEO Specialist and more. ### 2. It reaches for your skills, apps and servers ### 3. You approve what matters ## Built in from day one: the work a store needs done. Every store gets these tools on the first day. Your skills, apps and servers add to them. - **Builds your store**: Writes and edits your storefront code, runs the build, and checks each page before it goes live. - **Reads the web**: Clones a site you like, searches the web, takes screenshots and saves images. - **Makes images and video**: Generates product shots and short videos, and removes backgrounds. - **Runs your catalog**: Adds and updates products, imports a spreadsheet, and sets sale prices and promotions. - **Watches stock and orders**: Updates inventory, checks stock levels, and lists orders and customers. - **Sources products**: Imports dropshipping products and publishes print-on-demand items through Printful, Printify or another print partner. - **Publishes your site**: Deploys the store and connects your own domain. - **Reads your numbers**: Pulls analytics, charts what changed, and reports on SEO and how AI search sees your brand. - **Sets up payments**: Connects Stripe, and updates store settings and shipping rates. ## Skills teach it your trade, once. A skill is how you do a job, written down: your returns policy, your product copy rules, your launch checklist. The agent loads it when the job comes up and works your way every time. ## Your know-how, on file. Ready the moment the job comes up. Skills come with every plan, Free included. Choose which built-in agents use each one. ### Library: Start from twelve ready-made skills. Clone a website, add a wishlist, gift cards at checkout, product reviews, store SEO, deep research, product selection, booking products and more. Turn each one on or off for a store. ### Write your own: Write down how you do it. Plain words, plus any files the job needs. Give it a slash command and the phrases that should bring it up. Each save is a version you can restore, and Test it runs the skill before you rely on it. ### Import: Bring skills in from GitHub. Paste a link to a SKILL.md file, or to a repository that holds several skills. Runner adds them to your library. ## Connect the apps you already use. Sign in once. The agent finds the right app tool when a job needs it, and every tool follows the rule you set for the store. Featured apps: Gmail, Google Sheets, Slack, HubSpot, Notion, Google Calendar, Airtable, Perplexity. Plus the rest of the app catalog. Each connection belongs to the person who signed in. [See all integrations](https://www.runnerai.com/integrations) ### One rule per tool - **Allow**: Runs without asking. Good for tools that only read, like checking an inbox. - **Ask first**: Waits for your Approve in chat. A request you leave for 24 hours expires. - **Off**: The agent can't use it at all. Rules are set per store, so each store can trust the same app differently. ## Plug in your own MCP server. Run tools on your own systems, like a warehouse or a pricing sheet? Paste the server's URL. Runner finds its tools and asks you what each one may do. ## Safe by default. You loosen it, tool by tool. Up to 10 servers per store. Servers that speak MCP over HTTP. ### Reads on, changes ask, deletes off. Runner sorts every tool it finds and starts it safe. Tools the server adds later start as Ask first. ### Your key stays sealed. Choose no sign-in, or an API key sent in a header. The key is saved encrypted, and nobody can read it back. ### Only servers you trust. Paste the config from the server's docs and Runner fills in the form. You confirm you trust it, and pick which agents may use it. ## Build your own agent for the jobs you repeat. Start from a built-in agent or a blank page. Assign it to a chat when it's ready. ### Name and instructions Give it a name, a handle and a picture, and tell it how to work in your own words. ### Its own skills and apps Pick the skills, apps and MCP servers it may use. It uses only those. ### Its model Auto, Lite or Pro. On a plan without Pro, it runs on Lite. ### Drafts and versions Edit in a draft. Publishing makes a version, and a chat in progress picks it up from its next message. ### The Agent Marketplace List your agent for other stores, or add one someone else built and choose how it updates. ## Teach it your way. Let it do the work. ## Related - [Integrations](https://www.runnerai.com/integrations) - [Fly](https://www.runnerai.com/fly) - [Agent in chat](https://www.runnerai.com/agent-in-chat) - [Live](https://www.runnerai.com/live) - [Pricing](https://www.runnerai.com/pricing)