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

# Bring Your Own Agent

> Hubify works with any agent that speaks MCP, Claude Code, Codex, Cursor, Cline, Continue, Zed, Aider. Same skills, same data, your tool of choice.

# Bring Your Own Agent

Hubify Labs deliberately does NOT require the dedicated `hubify` CLI. The CLI is one nice path. The other path, the one most Captains actually use, is your existing AI coding agent.

Every skill in [Skills Overview](/skills/overview) is a markdown file plus a Hubify MCP server. If your agent reads MCP servers AND reads system-prompt-style markdown skills, you're done.

## What works today

| Agent                  | MCP support                          | Skills support                | Status                                                |
| ---------------------- | ------------------------------------ | ----------------------------- | ----------------------------------------------------- |
| **Claude Code**        | Native (`claude mcp add ...`)        | Native (`~/.claude/skills/`)  | First-class, [setup guide](/integrations/claude-code) |
| **Cursor**             | Native (`.cursor/mcp.json`)          | Via `.cursor/rules/*.mdc`     | First-class, [setup guide](/integrations/cursor)      |
| **Codex** (OpenAI CLI) | Via `~/.codex/config.json` MCP block | Via slash-command files       | Working, see below                                    |
| **Cline** (VS Code)    | Native MCP                           | Via custom-instructions field | Working                                               |
| **Continue**           | Native MCP                           | Via `.continue/rules`         | Working                                               |
| **Zed AI**             | Native MCP (assistant 2.0+)          | Via prompts library           | Working                                               |
| **Aider**              | None today                           | Via `--read` system prompt    | Manual, the MCP tools are unavailable                 |

## Codex (OpenAI CLI) setup

OpenAI's `codex` CLI ships MCP support. Add Hubify the same way you'd add any MCP server.

<Steps>
  <Step title="Install the CLI">
    ```bash theme={null}
    npm install -g hubify-labs
    ```

    Requires hubify-labs 2.0.0+ for the `hubify mcp` subcommand.
  </Step>

  <Step title="Edit ~/.codex/config.json">
    ```json theme={null}
    {
      "mcpServers": {
        "hubify": {
          "command": "hubify",
          "args": ["mcp"],
          "env": {
            "CONVEX_URL": "https://your-deployment.convex.cloud",
            "HUBIFY_LAB_SLUG": "your-lab-slug"
          }
        }
      }
    }
    ```
  </Step>

  <Step title="Verify">
    ```bash theme={null}
    codex
    > /hubify-status
    ```

    Should return your real lab status and recent events.
  </Step>
</Steps>

## Cline (VS Code) setup

<Steps>
  <Step title="Install the CLI">
    ```bash theme={null}
    npm install -g hubify-labs
    ```
  </Step>

  <Step title="Add to Cline MCP settings">
    Open Cline > Settings > MCP Servers and add:

    ```json theme={null}
    {
      "hubify": {
        "command": "hubify",
        "args": ["mcp"],
        "env": {
          "CONVEX_URL": "https://your-deployment.convex.cloud",
          "HUBIFY_LAB_SLUG": "your-lab-slug"
        }
      }
    }
    ```
  </Step>

  <Step title="Add skill prompts">
    Cline doesn't have a true slash-command system, but you can paste skill bodies into Custom Instructions and trigger them by phrase ("Run hubify-status").
  </Step>
</Steps>

## Continue setup

<Steps>
  <Step title="Install the CLI">
    ```bash theme={null}
    npm install -g hubify-labs
    ```
  </Step>

  <Step title="Edit ~/.continue/config.yaml">
    ```yaml theme={null}
    mcpServers:
      - name: hubify
        command: hubify
        args: ["mcp"]
        env:
          CONVEX_URL: https://your-deployment.convex.cloud
          HUBIFY_LAB_SLUG: your-lab-slug
    ```
  </Step>

  <Step title="Add skills to .continue/rules/">
    Continue reads `.continue/rules/*.md` as system-prompt rules. Drop skills there; trigger by phrase.
  </Step>
</Steps>

## Zed AI setup

<Steps>
  <Step title="Install the CLI">
    ```bash theme={null}
    npm install -g hubify-labs
    ```
  </Step>

  <Step title="Add to Zed settings">
    Open `~/.config/zed/settings.json`:

    ```json theme={null}
    {
      "context_servers": {
        "hubify": {
          "command": {
            "path": "hubify",
            "args": ["mcp"],
            "env": {
              "CONVEX_URL": "https://your-deployment.convex.cloud",
              "HUBIFY_LAB_SLUG": "your-lab-slug"
            }
          }
        }
      }
    }
    ```
  </Step>

  <Step title="Add skills to Zed prompts library">
    Open the prompts library (cmd-shift-p > "open prompts library") and paste skill bodies as new prompts.
  </Step>
</Steps>

## Aider (no MCP yet)

Aider does not support MCP. You can still wire some skills manually:

1. Install the CLI: `npm install -g hubify-labs`
2. Drop a system prompt: `aider --read ~/.aider/hubify-context.md` where the context file is the skill body plus `Use \`hubify\` CLI commands instead of MCP tools.\`
3. Aider will `hubify status`, `hubify experiments`, etc. shell out for the same data.

This is a workaround. If you want first-class Aider support, [open an issue](https://github.com/houstongolden/hubify/issues), we'll add a thin shim.

## Why this works

The whole architecture is intentional:

* **MCP server** = the data layer. Speaks one protocol (MCP). Every tool that adopts MCP gets Hubify for free.
* **Skills** = the prompt layer. Plain markdown. Trivial to port between agents.
* **Hubify Cloud** (Convex + Gateway + RunPod) = the source of truth. CLI, MCP, web app, and lab-site preview all read the same data.

You never have to choose between "use Claude Code" and "use Hubify." Both at once is the default.

## See also

* [Skills Overview](/skills/overview), what each skill does
* [Authoring a skill](/skills/authoring), write your own
* [MCP Server > Tools](/mcp/tools), the raw tool catalog skills build on
* [API Reference](/api/overview), REST surface (alternative to MCP for non-AI clients)