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

# Developer Platform Overview

> Choose the SketricGen MCP server, Admin API, Runtime API, or SDK for AI agent management, automation, streaming, and custom applications.

SketricGen exposes three developer surfaces. Choose the surface based on whether an AI assistant should manage SketricGen, your software should manage SketricGen resources, or your application should run an agent.

## Choose the right integration

| Surface                     | Use it when                                                                                         | Authentication                               | What it controls                                                                                       |
| --------------------------- | --------------------------------------------------------------------------------------------------- | -------------------------------------------- | ------------------------------------------------------------------------------------------------------ |
| **MCP server**              | Claude, ChatGPT, Codex, or another MCP client should build and manage Brand Agents conversationally | SketricGen OAuth or an Admin API key         | Brand Agent creation and editing, templates, knowledge bases, widget configuration, and connectors     |
| **Admin API**               | Your backend, CLI, or automation should manage Teamspace resources directly                         | Admin API key beginning with `sk_admin_`     | Projects, agents, Brand Agents, knowledge bases, conversations, traces, members, usage, and connectors |
| **Runtime API**             | Your application should send a message to an existing agent and receive its response                | Runtime API key beginning with `sk_runtime_` | Agent execution through `POST /api/v1/run-workflow`                                                    |
| **Python and Node.js SDKs** | You want a typed client for the Runtime API                                                         | Runtime API key beginning with `sk_runtime_` | Agent execution, streaming, and supported file uploads                                                 |

## How the surfaces fit together

```text theme={null}
AI assistant or MCP client
        |
        | Streamable HTTP + OAuth or sk_admin_ key
        v
SketricGen MCP server
        |
        | Calls the same authorized control plane
        v
SketricGen Admin API  <----- Your backend, CLI, or automation

Your product, website, or backend
        |
        | API-KEY: sk_runtime_...
        v
SketricGen Runtime API -----> Existing agent run and streamed response
```

The MCP server is an AI-friendly adapter over the Admin API. It does not bypass Teamspace roles, project boundaries, plan limits, or connector permissions.

The Runtime API is intentionally separate. An Admin API key cannot run an agent through the Runtime API, and a Runtime API key cannot create, edit, or delete Teamspace resources.

## Start here

* Connect an AI assistant with the [SketricGen MCP guide](/dev-guide/mcp-connect).
* Manage resources programmatically with the [Admin API](/dev-guide/admin-api).
* Run an existing agent with the [Runtime API](/dev-guide/public-api-usage).
* Use the [Python SDK](/dev-guide/python-sdk) or [Node.js SDK](/dev-guide/node-sdk) for a higher-level Runtime API client.

## API key location

Create API keys from the active Teamspace:

1. Open **Teamspace settings** in SketricGen.
2. Select **API keys**.
3. Select **Create key**.
4. Choose **Runtime key** or **Admin key** based on the integration.
5. Choose the narrowest project, agent, role, and expiration settings that support the task.
6. Copy the plaintext key immediately. SketricGen shows it only once.

API key creation requires an eligible Teamspace plan and permission to manage API keys.

## Security rule

Treat both key types as secrets:

* Keep keys on a server or in a secret manager.
* Never put a key in a public repository, screenshot, browser bundle, or prompt transcript.
* Prefer OAuth for hosted MCP clients because the AI client never needs your Admin API key.
* Use a narrowly scoped key for local agents and automation.
* Revoke a key from Teamspace settings when it is no longer needed.