Quickstart - AI Coding Tools

Pick a use case, prompt your coding agent, ship.

Give your coding agent the context it needs to start building on Corti. The fastest way is to connect the Corti MCP server once, then just describe what you want to build. If your tool doesn’t support MCP yet, or you’d rather point your agent at one specific use case, copy a prompt from the tabs below instead.

This guide uses AI coding agents to build applications on the Corti API. If instead you want to use Corti Models as the backend for your coding agent (OpenCode, ForgeCode, Crush, or Pi), see Use Corti Models in your coding agent.

1. Connect the Corti MCP (recommended)

   Add https://docs.corti.ai/mcp as an MCP server. Your agent can then browse Corti’s skills, SDK reference, and guides on its own and pull in whatever’s relevant to what you ask for — no need to hand it a specific skill URL.

   This is the Corti documentation MCP server. It’s unrelated to the MCP servers and connectors a Corti Agentic Framework app calls at runtime — see Connectors for that.

   Claude Code

   claude mcp add --transport http corti-docs https://docs.corti.ai/mcp
   

   Cursor

   Add to ~/.cursor/mcp.json (or .cursor/mcp.json in your project):

   {
     "mcpServers": {
       "corti-docs": {
         "url": "https://docs.corti.ai/mcp"
       }
     }
   }
   

   Codex

   codex mcp add corti-docs --url https://docs.corti.ai/mcp
   

   Then just prompt your agent, for example:

   Prompt

   Build a medical dictation web app with the Corti SDK.
   
   Credentials are in the Corti Console:
   https://console.corti.app → Developer Quickstart
   

   Lovable doesn’t support connecting MCP servers yet. Use the skill-based prompts in the next step instead.

2. Pick a use case and prompt your coding agent

   Use this if your tool doesn’t support MCP yet (like Lovable), or if you’d rather point your agent at one specific skill instead of connecting the full MCP server. Each prompt points your agent at a self-contained Corti skill — a single Markdown file with everything the agent needs to scaffold, wire auth, and run a working demo end-to-end.

   Dictation app

   Prompt

   Build a medical dictation web app with the Corti SDK.
   
   1. Fetch your build skill:
      https://docs.corti.ai/.well-known/agent-skills/corti-dictation/SKILL.md
   
   2. Credentials are in the Corti Console:
      https://console.corti.app → Developer Quickstart
   

   Ambient scribe

   Prompt

   Build an ambient scribe with the Corti SDK.
   
   1. Fetch your build skill:
      https://docs.corti.ai/.well-known/agent-skills/corti-ambient-scribe/SKILL.md
   
   2. Credentials are in the Corti Console:
      https://console.corti.app → Developer Quickstart
   

   Medical coding app

   Prompt

   Build a medical coding app with the Corti SDK.
   
   1. Fetch your build skill:
      https://docs.corti.ai/.well-known/agent-skills/corti-medical-coding/SKILL.md
   
   2. Credentials are in the Corti Console:
      https://console.corti.app → Developer Quickstart
   

   Chat assistant

   Prompt

   Build a chat assistant with the Corti Agentic Framework.
   
   1. Fetch your build skill:
      https://docs.corti.ai/.well-known/agent-skills/corti-agentic-assistant/SKILL.md
   
   2. Credentials are in the Corti Console:
      https://console.corti.app → Developer Quickstart
   

   Deeplinks open Claude Code, Cursor, or Codex with the prompt pre-filled. If a deeplink doesn’t launch (the app isn’t installed, or your browser blocks custom schemes), copy the prompt above and paste it into your tool directly.

3. Copy credentials into your app

   Your agent will ask you to add Corti credentials. Grab them from the Corti Console’s Developer Quickstart — it has a one-click “Copy all as .env variables” action.

   Open the Developer Quickstart in the Corti Console

   Sign in, then under Default client → Copy all as .env variables.

Other languages or custom integrations

The four skills above cover the most common use cases against the official JavaScript SDK (@corti/sdk) and the Corti Agentic Framework. For Python, Go, Ruby, or anything that doesn’t fit the four skills, point your agent at the machine-readable docs and let it build from there.

llms.txt

Concise API reference optimized for LLM context windows. Best for quick lookups and targeted code generation.

llms-full.txt

Complete documentation — guides, examples, and full specifications. Use when you need the whole picture.

Reference these URLs directly in your prompt (e.g. “Using https://docs.corti.ai/llms-full.txt, generate a Python client that …”), or add them to your project’s CLAUDE.md / AGENTS.md / Cursor Docs source so every session has the context.

Next steps

API Reference

Browse the full Corti API with interactive examples.

JavaScript SDK

SDK reference — what the skills build on under the hood.

Agentic Framework

Multi-expert orchestration, MCP servers, and the agent runtime.

Use Corti Models in your coding agent

Point OpenCode, ForgeCode, Crush, or Pi at Corti Models with the Corti CLI.

Example code

Reference implementations for transcription, ambient, coding, and agents.

AI-generated code should always be reviewed and tested before use in production. Human oversight ensures correctness, security, and compliance with applicable regulations.