@kapa.ai does Agent engine platform available for the self hosted open source airbyte?

Based on the knowledge sources, the Agent Engine platform is a cloud-hosted service (available at

app.airbyte.ai

) and is not the same as self-hosted open source Airbyte. Here's the distinction: • Agent Engine (hosted/platform mode): A cloud platform managed by Airbyte that handles connector management, credential storage, entity caching, and multi-tenant deployments. This requires an Airbyte subscription and credentials (

AIRBYTE_CLIENT_ID

+

AIRBYTE_CLIENT_SECRET

) from

app.airbyte.ai

. [About Agent Engine] • Open source agent connectors: If you don't subscribe to Airbyte's cloud platform, you can still use Airbyte's agent connectors as standalone Python packages directly in your app — without running the Airbyte platform at all. However, in this case, you are responsible for managing user credentials yourself, and you cannot use features like the context store. [Open source authentication] So, to directly answer your question: the Agent Engine platform (with managed credentials, UI, entity caching, etc.) is not available for self-hosted open source Airbyte — it is a cloud service. What is available for open source/self-hosted use are the agent connector Python packages, which you can import and use independently. [Agent connectors] 👍 0 button 👎 0 button ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.

@kapa.ai what is agent connectors, what is the difference between it with the previous connector? use gitlab connector as example

What Are Agent Connectors? Airbyte's agent connectors are lightweight Python packages that equip AI agents to call third-party APIs through strongly typed, well-documented tools. They are designed for real-time, on-demand data access within AI agent applications — not for batch data movement. [Agent connectors] Agent Connectors vs. Traditional (Replication) Connectors [Agent connectors] GitLab Connector as an Example The GitLab agent connector (

airbyte-agent-gitlab

) connects to the GitLab REST API (v4) and gives AI agents real-time access to projects, issues, merge requests, commits, pipelines, groups, branches, releases, tags, members, milestones, and users. [GitLab connector] What it can do (supported entities & actions): • Projects, Issues, Merge Requests, Users, Commits, Groups, Branches, Pipelines, Releases, Tags, Milestones, Members — all support

List

,

Get

, and

Search

actions. What it cannot do (unsupported): • Create a new project, delete an issue, merge a merge request, or trigger a pipeline — write operations are not currently supported. [GitLab connector] Installation

uv pip install airbyte-agent-gitlab

Usage (Open Source mode)

from airbyte_agent_gitlab import GitlabConnector
from airbyte_agent_gitlab.models import GitlabPersonalAccessTokenAuthConfig

connector = GitlabConnector(
    auth_config=GitlabPersonalAccessTokenAuthConfig(
        access_token="<your_gitlab_personal_access_token>"
    )
)

@agent.tool_plain  # assumes you're using Pydantic AI
@GitlabConnector.tool_utils
async def gitlab_execute(entity: str, action: str, params: dict | None = None):
    return await connector.execute(entity, action, params or {})

[GitLab connector] Usage (Hosted mode — credentials managed by Airbyte Cloud)

from airbyte_agent_gitlab import GitlabConnector, AirbyteAuthConfig

connector = GitlabConnector(
    auth_config=AirbyteAuthConfig(
        customer_name="<your_customer_name>",
        organization_id="<your_organization_id>",  # Optional for multi-org clients
        airbyte_client_id="<your-client-id>",
        airbyte_client_secret="<your-client-secret>"
    )
)

@agent.tool_plain
@GitlabConnector.tool_utils
async def gitlab_execute(entity: str, action: str, params: dict | None = None):
    return await connector.execute(entity, action, params or {})

[GitLab connector] In summary: agent connectors complement (not replace) traditional replication connectors. Use agent connectors when your AI agent needs live, on-demand data access. Use replication connectors when you need scheduled, bulk data movement into a warehouse. 👍 0 button 👎 0 button ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help., with interactive elements

@kapa.ai ok, got it. this is not what I'm looking for. so if we have a ai platform, we would like to enable it with ability to create connections on our hosted airbyte, is there any mcp server support this or we need to host our own mcp?

Based on the knowledge sources, there are two relevant MCP server options, but with important caveats regarding your specific use case (managing connections on a self-hosted Airbyte instance): 1. Airbyte Agent Engine MCP (Hosted — mcp.airbyte.ai) This is a fully hosted, remote MCP server that requires no installation. However, it connects to Airbyte's cloud platform (

app.airbyte.ai

), not your self-hosted Airbyte instance. [Agent Engine MCP] 2. PyAirbyte MCP Server (Local — Self-Hosted) This is a local MCP server you run yourself, and it is more relevant to your use case. It provides tools to: • List available Airbyte connectors • Validate configurations • Run sync operations • Manage connections in your Airbyte Cloud workspace (via

AIRBYTE_CLOUD_CLIENT_ID

,

AIRBYTE_CLOUD_CLIENT_SECRET

,

AIRBYTE_CLOUD_WORKSPACE_ID

) You run it using

uvx

with the following config:

{
  "mcpServers": {
    "airbyte": {
      "command": "uvx",
      "args": [
        "--python=3.11",
        "--from=airbyte@latest",
        "airbyte-mcp"
      ],
      "env": {
        "AIRBYTE_MCP_ENV_FILE": "/path/to/my/.mcp/airbyte_mcp.env",
        "AIRBYTE_CLOUD_MCP_SAFE_MODE": "1",
        "AIRBYTE_CLOUD_MCP_READONLY_MODE": "0"
      }
    }
  }
}

[PyAirbyte MCP] Important Limitation The knowledge sources only mention the PyAirbyte MCP connecting to Airbyte Cloud (via cloud credentials), not to a self-hosted open source Airbyte instance. Whether it supports connecting to a self-hosted Airbyte via its REST API is not explicitly stated in the available documentation. If your goal is specifically to manage connections on a self-hosted Airbyte, you may need to explore the PyAirbyte GitHub Issues or Discussions to confirm whether self-hosted Airbyte is supported. [PyAirbyte MCP] 👍 0 button 👎 0 button ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.