Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
53 changes: 53 additions & 0 deletions content/docs/mcp_servers/gpt_image_acedata.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,53 @@
---
title: GPT Image by Ace Data Cloud
icon: Plug
description: Install GPT Image image tools in LibreChat using an independent Ace Data Cloud Agent Plugin and a personal MCP connection.
---

This independent integration connects **GPT Image** through **Ace Data Cloud**. It is not published by the GPT Image service brand or the LibreChat maintainers. Model-chat credentials and this tool connection are separate.

## Install the native package

The experimental [Agent Plugins](/docs/features/agent_plugins) format loads each operator-installed plugin at startup. Review the [package source](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0/plugins/gpt-image) and install only the service you need:

```bash
python3 scripts/install.py gpt-image --plugin-dir /path/to/LibreChat/plugin
```

[Release v0.1.0](https://github.com/AceDataCloud/LibreChatPlugins/releases/tag/v0.1.0) contains one independently downloadable ZIP per service plus `SHA256SUMS`. Verify the selected file, then extract its single service directory into the administrator-controlled plugin directory; preserve existing plugin/data directories. Each archive includes its own license, English/Chinese illustrated guides and screenshots. Alternatively, run the command above from a `v0.1.0` source checkout of `AceDataCloud/LibreChatPlugins`. The selected plugin directory must be visible inside the LibreChat filesystem (`/app/plugin` by default); preserve existing Docker mounts. Restart that instance. The package adds the `gpt-image-by-acedatacloud` MCP connection and a service-specific Skill to that instance's tool catalog. It is not an installation into every LibreChat instance or a global marketplace entry.

The [English and Chinese illustrated package guides](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0) include installation, account setup, current limits and troubleshooting. The package reuses [OpenAIMCP](https://github.com/AceDataCloud/OpenAIMCP); it does not implement its own billing or generation API.

## Connect your account

![Official LibreChat MCP settings reference](/images/mcp/mcp_ui_settings_panel.png)

This official screenshot shows LibreChat v0.8.2 and a different sample server. It locates the settings controls; it is not an Ace connection or a guarantee of the current UI layout.


1. Register or sign in at [Ace Data Cloud Applications](https://platform.acedata.cloud/console/applications?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=gpt-image_official_guide_key).
2. Find **General application**, or initialize it if absent. Check GPT Image access, [current pricing and balance](https://platform.acedata.cloud/models?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=gpt-image_official_guide_pricing).
3. For the native URL-only plugin, open **+ → MCP Servers** or **MCP Settings**, select `gpt-image-by-acedatacloud`, and complete the Ace OAuth connection in the same browser session. A discovered service is not yet an authenticated connection.

For an explicit personal Token route, use the separately configured [customUserVars](/docs/configuration/librechat_yaml/object_structure/mcp_servers#customuservars) fields instead of putting credentials in a plugin. Copy the application API key or choose **Keys / Manage Keys → Create**, set the service/API restrictions and expiry you need, and paste the raw value in the service's MCP Token field. Do not use an account-management Token or the chat-model key dialog. A YAML MCP entry takes precedence over a same-name plugin server; choose one authentication route.

![Copy an API key or open Manage Keys](/images/mcp/acedata_get_api_key.png)

![Create a separately named key](/images/mcp/acedata_create_api_key.png)

These are real Ace Console screenshots with redacted values, reused from the [pinned onboarding source](https://github.com/AceDataCloud/SunoDify/tree/8d3d03570e73324220f111f7fe343da0792a8071/_assets/tutorial). Their original capture date is not specified here; they are not LibreChat execution evidence. Account credit and access depend on the current account.

## First use and task results

Confirm the connection lists tools before sending a paid operation. To inspect an existing result, call `openai_get_task` with a task ID already belonging to your account. If the host requires a model to trigger the call, that model can have its own cost.

For a new `openai_generate_image` request, review the exact tool schema and current price first. Submit once after the user confirms the selected service, inputs and bounded spend. Store the returned task ID and query the same ID until terminal success or failure. Do not regenerate to poll. A pending preview is not final media, and a connected plugin does not prove a unique bill. Stop on terminal failure.

## Troubleshooting

- Missing service: check the host implements Agent Plugins, directory mounting and startup diagnostics.
- Authentication error: verify the chosen OAuth or personal-Token route, API permissions, expiry and balance. Do not combine fixed credentials with OAuth on one connection.
- Pending: retrieve the same task ID; do not repeat generation.
- Rate limit or timeout: inspect the existing response/task and reduce concurrency before a new request.

The package uses the existing OpenAI MCP, which exposes additional OpenAI tools. Its GPT Image Skill is guidance, not an authorization boundary; restrict the account key to the APIs needed.
53 changes: 53 additions & 0 deletions content/docs/mcp_servers/happyhorse_acedata.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,53 @@
---
title: HappyHorse by Ace Data Cloud
icon: Plug
description: Install HappyHorse video tools in LibreChat using an independent Ace Data Cloud Agent Plugin and a personal MCP connection.
---

This independent integration connects **HappyHorse** through **Ace Data Cloud**. It is not published by the HappyHorse service brand or the LibreChat maintainers. Model-chat credentials and this tool connection are separate.

## Install the native package

The experimental [Agent Plugins](/docs/features/agent_plugins) format loads each operator-installed plugin at startup. Review the [package source](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0/plugins/happyhorse) and install only the service you need:

```bash
python3 scripts/install.py happyhorse --plugin-dir /path/to/LibreChat/plugin
```

[Release v0.1.0](https://github.com/AceDataCloud/LibreChatPlugins/releases/tag/v0.1.0) contains one independently downloadable ZIP per service plus `SHA256SUMS`. Verify the selected file, then extract its single service directory into the administrator-controlled plugin directory; preserve existing plugin/data directories. Each archive includes its own license, English/Chinese illustrated guides and screenshots. Alternatively, run the command above from a `v0.1.0` source checkout of `AceDataCloud/LibreChatPlugins`. The selected plugin directory must be visible inside the LibreChat filesystem (`/app/plugin` by default); preserve existing Docker mounts. Restart that instance. The package adds the `happyhorse-by-acedatacloud` MCP connection and a service-specific Skill to that instance's tool catalog. It is not an installation into every LibreChat instance or a global marketplace entry.

The [English and Chinese illustrated package guides](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0) include installation, account setup, current limits and troubleshooting. The package reuses [HappyHorseMCP](https://github.com/AceDataCloud/HappyHorseMCP); it does not implement its own billing or generation API.

## Connect your account

![Official LibreChat MCP settings reference](/images/mcp/mcp_ui_settings_panel.png)

This official screenshot shows LibreChat v0.8.2 and a different sample server. It locates the settings controls; it is not an Ace connection or a guarantee of the current UI layout.


1. Register or sign in at [Ace Data Cloud Applications](https://platform.acedata.cloud/console/applications?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=happyhorse_official_guide_key).
2. Find **General application**, or initialize it if absent. Check HappyHorse access, [current pricing and balance](https://platform.acedata.cloud/models?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=happyhorse_official_guide_pricing).
3. For the native URL-only plugin, open **+ → MCP Servers** or **MCP Settings**, select `happyhorse-by-acedatacloud`, and complete the Ace OAuth connection in the same browser session. A discovered service is not yet an authenticated connection.

For an explicit personal Token route, use the separately configured [customUserVars](/docs/configuration/librechat_yaml/object_structure/mcp_servers#customuservars) fields instead of putting credentials in a plugin. Copy the application API key or choose **Keys / Manage Keys → Create**, set the service/API restrictions and expiry you need, and paste the raw value in the service's MCP Token field. Do not use an account-management Token or the chat-model key dialog. A YAML MCP entry takes precedence over a same-name plugin server; choose one authentication route.

![Copy an API key or open Manage Keys](/images/mcp/acedata_get_api_key.png)

![Create a separately named key](/images/mcp/acedata_create_api_key.png)

These are real Ace Console screenshots with redacted values, reused from the [pinned onboarding source](https://github.com/AceDataCloud/SunoDify/tree/8d3d03570e73324220f111f7fe343da0792a8071/_assets/tutorial). Their original capture date is not specified here; they are not LibreChat execution evidence. Account credit and access depend on the current account.

## First use and task results

Confirm the connection lists tools before sending a paid operation. To inspect an existing result, call `happyhorse_get_task` with a task ID already belonging to your account. If the host requires a model to trigger the call, that model can have its own cost.

For a new `happyhorse_generate_video` request, review the exact tool schema and current price first. Submit once after the user confirms the selected service, inputs and bounded spend. Store the returned task ID and query the same ID until terminal success or failure. Do not regenerate to poll. A pending preview is not final media, and a connected plugin does not prove a unique bill. Stop on terminal failure.

## Troubleshooting

- Missing service: check the host implements Agent Plugins, directory mounting and startup diagnostics.
- Authentication error: verify the chosen OAuth or personal-Token route, API permissions, expiry and balance. Do not combine fixed credentials with OAuth on one connection.
- Pending: retrieve the same task ID; do not repeat generation.
- Rate limit or timeout: inspect the existing response/task and reduce concurrency before a new request.


15 changes: 15 additions & 0 deletions content/docs/mcp_servers/index.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -16,4 +16,19 @@ Use these guides to configure specific MCP servers with the right transport, OAu
<Cards.Card title="MCP overview" href="/docs/features/mcp" arrow>
Learn how MCP works in LibreChat before adding product-specific servers.
</Cards.Card>
<Cards.Card title="Suno by Ace Data Cloud" href="/docs/mcp_servers/suno_acedata" arrow>
Independent music tools, account connection, and same-task retrieval.
</Cards.Card>
<Cards.Card title="Seedance by Ace Data Cloud" href="/docs/mcp_servers/seedance_acedata" arrow>
Independent video tools, account connection, and same-task retrieval.
</Cards.Card>
<Cards.Card title="HappyHorse by Ace Data Cloud" href="/docs/mcp_servers/happyhorse_acedata" arrow>
Independent video tools, account connection, and same-task retrieval.
</Cards.Card>
<Cards.Card title="GPT Image by Ace Data Cloud" href="/docs/mcp_servers/gpt_image_acedata" arrow>
Independent image tools, account connection, and same-task retrieval.
</Cards.Card>
<Cards.Card title="Nano Banana by Ace Data Cloud" href="/docs/mcp_servers/nano_banana_acedata" arrow>
Independent image tools, account connection, and same-task retrieval.
</Cards.Card>
</Cards>
11 changes: 10 additions & 1 deletion content/docs/mcp_servers/meta.json
Original file line number Diff line number Diff line change
@@ -1,5 +1,14 @@
{
"title": "MCP Servers",
"icon": "Blocks",
"pages": ["index", "google_workspace", "salesforce"]
"pages": [
"index",
"google_workspace",
"salesforce",
"suno_acedata",
"seedance_acedata",
"happyhorse_acedata",
"gpt_image_acedata",
"nano_banana_acedata"
]
}
53 changes: 53 additions & 0 deletions content/docs/mcp_servers/nano_banana_acedata.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,53 @@
---
title: Nano Banana by Ace Data Cloud
icon: Plug
description: Install Nano Banana image tools in LibreChat using an independent Ace Data Cloud Agent Plugin and a personal MCP connection.
---

This independent integration connects **Nano Banana** through **Ace Data Cloud**. It is not published by the Nano Banana service brand or the LibreChat maintainers. Model-chat credentials and this tool connection are separate.

## Install the native package

The experimental [Agent Plugins](/docs/features/agent_plugins) format loads each operator-installed plugin at startup. Review the [package source](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0/plugins/nano-banana) and install only the service you need:

```bash
python3 scripts/install.py nano-banana --plugin-dir /path/to/LibreChat/plugin
```

[Release v0.1.0](https://github.com/AceDataCloud/LibreChatPlugins/releases/tag/v0.1.0) contains one independently downloadable ZIP per service plus `SHA256SUMS`. Verify the selected file, then extract its single service directory into the administrator-controlled plugin directory; preserve existing plugin/data directories. Each archive includes its own license, English/Chinese illustrated guides and screenshots. Alternatively, run the command above from a `v0.1.0` source checkout of `AceDataCloud/LibreChatPlugins`. The selected plugin directory must be visible inside the LibreChat filesystem (`/app/plugin` by default); preserve existing Docker mounts. Restart that instance. The package adds the `nano-banana-by-acedatacloud` MCP connection and a service-specific Skill to that instance's tool catalog. It is not an installation into every LibreChat instance or a global marketplace entry.

The [English and Chinese illustrated package guides](https://github.com/AceDataCloud/LibreChatPlugins/tree/v0.1.0) include installation, account setup, current limits and troubleshooting. The package reuses [NanoBananaMCP](https://github.com/AceDataCloud/NanoBananaMCP); it does not implement its own billing or generation API.

## Connect your account

![Official LibreChat MCP settings reference](/images/mcp/mcp_ui_settings_panel.png)

This official screenshot shows LibreChat v0.8.2 and a different sample server. It locates the settings controls; it is not an Ace connection or a guarantee of the current UI layout.


1. Register or sign in at [Ace Data Cloud Applications](https://platform.acedata.cloud/console/applications?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=nano-banana_official_guide_key).
2. Find **General application**, or initialize it if absent. Check Nano Banana access, [current pricing and balance](https://platform.acedata.cloud/models?utm_source=librechat&utm_medium=referral&utm_campaign=evergreen&utm_content=nano-banana_official_guide_pricing).
3. For the native URL-only plugin, open **+ → MCP Servers** or **MCP Settings**, select `nano-banana-by-acedatacloud`, and complete the Ace OAuth connection in the same browser session. A discovered service is not yet an authenticated connection.

For an explicit personal Token route, use the separately configured [customUserVars](/docs/configuration/librechat_yaml/object_structure/mcp_servers#customuservars) fields instead of putting credentials in a plugin. Copy the application API key or choose **Keys / Manage Keys → Create**, set the service/API restrictions and expiry you need, and paste the raw value in the service's MCP Token field. Do not use an account-management Token or the chat-model key dialog. A YAML MCP entry takes precedence over a same-name plugin server; choose one authentication route.

![Copy an API key or open Manage Keys](/images/mcp/acedata_get_api_key.png)

![Create a separately named key](/images/mcp/acedata_create_api_key.png)

These are real Ace Console screenshots with redacted values, reused from the [pinned onboarding source](https://github.com/AceDataCloud/SunoDify/tree/8d3d03570e73324220f111f7fe343da0792a8071/_assets/tutorial). Their original capture date is not specified here; they are not LibreChat execution evidence. Account credit and access depend on the current account.

## First use and task results

Confirm the connection lists tools before sending a paid operation. To inspect an existing result, call `nanobanana_get_task` with a task ID already belonging to your account. If the host requires a model to trigger the call, that model can have its own cost.

For a new `nanobanana_generate_image` request, review the exact tool schema and current price first. Submit once after the user confirms the selected service, inputs and bounded spend. Store the returned task ID and query the same ID until terminal success or failure. Do not regenerate to poll. A pending preview is not final media, and a connected plugin does not prove a unique bill. Stop on terminal failure.

## Troubleshooting

- Missing service: check the host implements Agent Plugins, directory mounting and startup diagnostics.
- Authentication error: verify the chosen OAuth or personal-Token route, API permissions, expiry and balance. Do not combine fixed credentials with OAuth on one connection.
- Pending: retrieve the same task ID; do not repeat generation.
- Rate limit or timeout: inspect the existing response/task and reduce concurrency before a new request.


Loading