> ## Documentation Index
> Fetch the complete documentation index at: https://developer.eka.care/llms.txt
> Use this file to discover all available pages before exploring further.

# Build with AI

> Set up a coding agent to build with EkaScribe: the reference skill and the CLAUDE.md file.

Use these to have a coding assistant build your EkaScribe integration. Each one hands it the same facts as these docs.

> [!NOTE]
>
> Check what the assistant writes against these docs, and test it in the sandbox before it reaches a patient.

> **[Reference skill](#install-the-skill)**
>
> Every operation, its hosts, auth and headers, in one install.

> **[CLAUDE.md file](https://developer.eka.care/docs/scribe/build-with-ai/claude-md.md)**
>
> The four calls and their rules, in one file to drop into your project.

## Before you start

Three things have to be in place before anything on this page works.

| What you need | Why |
| - | - |
| A coding assistant | Claude Code, Cursor, Codex or ChatGPT. Any one is enough. Pick one and stay on it. |
| A paid plan on it | Free tiers stop partway through a long task. |
| Your own project open in it | The assistant reads and writes the folder it is opened in. Open the repository you are building in first. |

Then work down this page in order: install the skill, connect the MCP server where there is one, and ask for one flow at a time in plain language. Read what the assistant writes before you run it.

## Install the skill

The `eka-scribe` skill carries the hosts, auth and headers, and every Scribe operation, each with a link to its page here. It is built from these docs, so it changes when they do.

In Claude Code:

```bash theme={null}
/plugin marketplace add eka-care/eka-docs
/plugin install eka-scribe@eka-docs
```

In Cursor, Codex or another agent, copy the [skill folder](https://github.com/eka-care/eka-docs/tree/main/plugins/eka-scribe/skills/eka-scribe) into the folder your agent loads skills from. Its map is [`SKILL.md`](https://raw.githubusercontent.com/eka-care/eka-docs/main/plugins/eka-scribe/skills/eka-scribe/SKILL.md).

## Prompting an agent for Scribe

Tell the assistant these, and check its output against them.

| Tell it | Because |
| - | - |
| Keep the client secret on a server | The secret never goes in browser code. Exchange it on a small backend and hand the UI only the short-lived access token. |
| Upload raw audio bytes | The upload takes the bytes themselves: not multipart, not JSON, not base64. |
| Read the HTTP status when polling | 202 means still processing, 200 means done, 206 means some templates failed, and 410 means the session expired. |
| Use a chunked session for files over 10 MB | A single upload takes one file up to 10 MB. Longer recordings upload in numbered chunks before the session ends. |
| Sign in again on a 401 | The access token expires. A 401 means fetch a new one, not that the call is wrong. |

The [CLAUDE.md file](https://developer.eka.care/docs/scribe/build-with-ai/claude-md.md) carries these rules and the four calls in one file you can drop into a project.

## Read these docs as an agent

Every page here is also plain text an agent can read.

* [`llms.txt`](https://developer.eka.care/llms.txt) is a map of every page.
* [`llms-full.txt`](https://developer.eka.care/llms-full.txt) is every page's text in one file.
* The menu at the top of each page copies it as Markdown, or opens it in Claude or ChatGPT.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.