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

# EvoLink CLI / MCP

> Choose a route, install by client, verify access, and find complete references

Use EvoLink from your agent to generate images, video, music, and speech, discover models and costs, upload references, track tasks, and save originals. Choose a connection method, then follow its installation and free verification guide.

## Choose a connection method

| Where you work | Recommended route | Installation and authentication |
| - | - | - |
| Claude chat or ChatGPT | [Remote MCP](/docs/en/mcp/overview) | Add a connection and authorize browser OAuth; no CLI installation |
| Codex, Claude Code, Cursor, OpenClaw, or Hermes with terminal execution | [CLI](/docs/en/cli/setup) | Install the CLI, sign in, and install/load the skill |
| Terminal, SSH, or scripts | [CLI](/docs/en/cli/setup) | Install in the execution environment; SSH has separate callback steps |
| A terminal agent using native MCP tools | [MCP client guide](/docs/en/mcp/overview#choose-your-assistant) | Configure and authorize the host connection directly |
| A client supporting only local stdio | [Local MCP](/docs/en/mcp/other-clients#local-mcp) | Configure a local process, Node.js, and a personal API Key |

Codex CLI is your agent client; EvoLink CLI is the media tool it calls. They are different products. Each card recommends one route; CLI client guides also link to the native MCP alternative.

## Choose your client

<CardGroup cols={2}>
  <Card title="Claude" icon="https://cdn.evolink.ai/mcp/clients/claude.webp" href="/docs/en/mcp/claude">Remote MCP · browser sign-in</Card>
  <Card title="Codex" icon="https://cdn.evolink.ai/mcp/clients/codex.svg" href="/docs/en/cli/codex">CLI · install where Codex runs</Card>
  <Card title="Claude Code" icon="https://mintcdn.com/muyutechnology/nT64SbOJQf4nX61W/images/mcp/claude-code.svg?fit=max&auto=format&n=nT64SbOJQf4nX61W&q=85&s=4bbd9101fa0b0a0eb4640ff3e3574865" href="/docs/en/cli/claude-code" width="24" height="24" data-path="images/mcp/claude-code.svg">CLI · install the command and companion skill</Card>
  <Card title="ChatGPT" icon="https://cdn.evolink.ai/mcp/clients/openai.svg" href="/docs/en/mcp/chatgpt">Remote MCP · add and authorize a connection</Card>
  <Card title="Cursor" icon="https://cdn.evolink.ai/mcp/clients/cursor.webp" href="/docs/en/cli/cursor">CLI · use the terminal available to your agent</Card>
  <Card title="OpenClaw" icon="https://cdn.evolink.ai/mcp/clients/openclaw.svg" href="/docs/en/cli/openclaw">CLI · install in the Gateway execution environment</Card>
  <Card title="Hermes" icon="https://cdn.evolink.ai/mcp/clients/hermes.png" href="/docs/en/cli/hermes-agent">CLI · install where Hermes executes commands</Card>
  <Card title="EvoLink CLI" icon="terminal" href="/docs/en/cli/setup">Detailed installation for terminals, SSH, and scripts</Card>
  <Card title="Other MCP clients" icon="plug" href="/docs/en/mcp/other-clients#remote-mcp">Generic Streamable HTTP and OAuth</Card>
  <Card title="Local MCP" icon="key" href="/docs/en/mcp/other-clients#local-mcp">stdio · run MCP tools locally</Card>
</CardGroup>

## Verify before generating

<Steps>
  <Step title="Install and authenticate your chosen route">
    Follow the five-step [CLI installation](/docs/en/cli/setup) or your [MCP client guide](/docs/en/mcp/overview#choose-your-assistant). Complete browser sign-in yourself. Keep Keys and tokens out of chat.
  </Step>

  <Step title="Query balance and models, and verify skills or tools">
    CLI users query balance, run doctor, and ask the agent to confirm skill discovery. MCP users actually call check\_balance and search\_models. Installed files or a tool list do not establish successful execution. These checks do not submit paid media tasks.
  </Step>

  <Step title="Review the plan and estimated cost before approving generation">
    Check the exact model and inputs, prepare references, and review the plan and estimate. After explicit approval, submit once, retain the task ID, and save originals. See [CLI workflows](/docs/en/cli/workflows) or [MCP workflows](/docs/en/mcp/overview#workflows).
  </Step>
</Steps>

## How the routes relate

CLI media requests follow `Agent / terminal → CLI → platform API`. Native MCP follows `Agent → remote or local MCP → platform API`. CLI media execution does not depend on remote MCP. Sign-in uses Passport; uploads and original downloads have separate file endpoints.

Installation, sessions, and connection checks are separate. Same-account OAuth can currently share internal Key permissions and quotas. Use one route for the task. Recover unknown submissions through the original route rather than generating again through the other.

| What to read next | Reference |
| - | - |
| CLI commands, options, JSON, and API mapping | [CLI reference](/docs/en/cli/reference) |
| MCP tools, inputs, results, and recovery | [MCP reference](/docs/en/mcp/tools) |
| Model selection, estimates, budgets, and access | [Models, costs, and quotas](/docs/en/cli-mcp/models-and-billing) |
| Task states, uploads, downloads, and expiry | [Tasks, references, and results](/docs/en/cli-mcp/tasks-and-files) |
| Skills, plugins, Markdown, and llms.txt | [Agent resources](/docs/en/cli-mcp/agent-resources) |
| Installation or connection failures | [CLI troubleshooting](/docs/en/cli/billing#faq) / [MCP troubleshooting](/docs/en/mcp/billing#faq) |

These guides describe released CLI 0.8.1 and MCP 1.6.1. Development candidates are documented after release acceptance. Query the connected service for current tools and account-available models.


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