# Z.ai MCP connector

Connect Z.ai to Elaichi and Claude, ChatGPT, Cursor, or the Elaichi Agent can generate images and video, transcribe audio, parse documents, search the web, and run Z.ai agents, with every request logged under the person who made it.

Source: https://elaichi.ai/connectors/zai/

## Facts

| | |
| --- | --- |
| Application | Z.ai |
| Category | Artificial Intelligence |
| AI tools | 14 |
| Authentication | Connects with an API key |
| Needs your own OAuth app | No |
| MCP endpoint | https://api.elaichi.ai/mcp |
| Works with | Claude, ChatGPT, Cursor, any MCP client, and the Elaichi Agent |
| Tools advertised by name | Yes |

## What you can ask once Z.ai is connected

- Summarize this week's Z.ai image generations by team.
- Which transcriptions did Support run in Z.ai yesterday?
- List Z.ai files uploaded by Research this month.

## Connect Z.ai in Elaichi

This happens once for the organization, before any client is involved.

1. Open Connections, choose Add connection, and pick Z.ai.
2. Optionally set Share with, then press Connect.
3. Paste a Z.ai API key. One person generates a token in Z.ai and pastes it once. Everyone else works through Share with, and never sees it.

Credentials are vaulted and nobody, including the AI, reads them back. The connection becomes a toolbox immediately, so you can curate which Z.ai tools are exposed, rename them, or freeze arguments before anyone points a client at it.

## Connect Z.ai to Claude

Endpoint: https://api.elaichi.ai/mcp

1. Open Customize, then Connectors.
2. Press Add.
3. Name it, paste the MCP server URL, then Continue.
4. Sign in and approve.

On Team and Enterprise, an Owner adds it once. Everyone else turns it on for themselves.

## Connect Z.ai to ChatGPT

Endpoint: https://api.elaichi.ai/mcp

1. Open Plugins, then press the + button.
2. Name it and paste the endpoint into Server URL.
3. Leave Authentication on OAuth, then tick the risk acknowledgement.
4. Press Create, then sign in and approve.

Works on the web today. The plugin directory lives at chatgpt.com/plugins.

## Connect Z.ai to Cursor

Endpoint: https://api.elaichi.ai/mcp

1. Open `~/.cursor/mcp.json`.
2. Add the endpoint under `mcpServers`.
3. Reload Cursor, then sign in and approve.

Set up per machine, so repeat it on each computer you work from.

## Connect Z.ai to any MCP client

Endpoint: https://api.elaichi.ai/mcp

1. Add the endpoint as a remote MCP server.
2. Sign in and approve.

The Elaichi Agent already has these tools, with nothing to set up.

## What the consent screen decides

Only Read is granted by default, which is not enough to call a Z.ai tool. Over MCP there is no trusted place to confirm a write in the moment, so the consent screen is the standing approval rather than a formality. Grant Read and Run tools. Think hard before granting Delete, which reaches into connected apps and cannot be undone.

## What teams do with Z.ai through Elaichi

### Turn a campaign brief into visuals

Marketing. Describe the shot you need and get Z.ai images back for a landing page, ad, or social post. Ask for a different angle or color palette and the next version arrives in the same conversation.

### Draft a short product video from a script

Marketing. Hand Z.ai a scene description and a script, then check back for the finished clip once it has rendered. Longer jobs run in the background so nobody sits and waits.

### Transcribe a recorded customer call

Support. Drop in the recording and Z.ai returns a transcript you can read, quote from, and turn into follow-up actions. No more scrubbing through audio to find the moment a customer raised a complaint.

### Pull the tables out of a scanned contract

Operations. Upload a PDF or scan and Z.ai's layout parsing returns the headings, paragraphs, and tables in order. Ask for the payment terms or renewal dates and get them from the right section.

### Read a competitor page and brief the team

Research. Ask for a web search on a topic, have Z.ai read the pages that matter, and get a summary with the sources named. Useful before a pitch, a board update, or a pricing review.

### Set up a Z.ai agent for a recurring task

Product. Create a Z.ai agent with the instructions for a job your team repeats, then start conversations with it and collect the results as they finish. One person configures it, everyone uses it.

## Frequently asked questions

### How do I connect Z.ai to Claude?

First connect Z.ai in Elaichi by pasting in the API key from your Z.ai account, which takes under a minute. There is no OAuth application to register and no client ID or secret to generate. Then in Claude open Customize, then Connectors, then Add, and paste https://api.elaichi.ai/mcp as the endpoint. Sign in with your Elaichi account and Z.ai appears in Claude ready to use.

### Does Z.ai work with ChatGPT and Cursor as well as Claude?

Yes. Once Z.ai is connected in Elaichi, the same endpoint, https://api.elaichi.ai/mcp, works in Claude, ChatGPT, Cursor, the Elaichi Agent, and any other MCP client. You connect Z.ai once and every client your team uses picks it up.

### What can an AI agent actually do with Z.ai?

Through this connector an agent can ask Z.ai to write or answer in chat, generate images and video, transcribe audio recordings, and parse the layout of a PDF or scanned document into readable text and tables. It can also run a Z.ai web search, read a web page into text, upload a file for Z.ai to work on, count tokens before a long request, and create and talk to Z.ai agents. Long jobs like video run in the background and the agent collects the result when it is ready.

### Does connecting Z.ai give the AI access to everything in my Z.ai account?

Z.ai connects with an API key, so the connection can do exactly what that key's Z.ai account allows and nothing beyond it. Elaichi can narrow that further, for example by allowing transcription and document parsing but not video generation, and it can never widen what the key already permits. Every request is recorded against the person who made it, not against the key.

### Can my team share one Z.ai connection?

Yes. One person connects Z.ai in Elaichi and shares the connection with a team, and nobody else ever sees or handles the API key. Each teammate still signs in to Elaichi as themselves, so the audit log names the person who generated an image or transcribed a call, not just the shared account.

### Can I stop an agent from running expensive or unwanted actions in Z.ai?

Yes. Restrictions in Elaichi apply per action, so you can allow Z.ai chat and transcription while blocking image or video generation for a given team. A blocked action is never shown to Claude, ChatGPT, Cursor, or any other client, so no prompt, however worded, can reach it.

### What happens to a Z.ai connection when someone leaves?

Offboarding a person in Elaichi ends their access to Z.ai at once across every client they had connected. If they were using a shared Z.ai connection, it keeps working for everyone else on the team. Disconnecting Z.ai once in Elaichi removes it from Claude, ChatGPT, Cursor, and every other client at the same time.

## All 14 Z.ai tools

Every tool below is callable through https://api.elaichi.ai/mcp once Z.ai is connected, subject to the toolbox it is in and the restrictions on the caller.

- **Create a Z.ai chat completion** (Create). Create a chat completion in zai that generates AI replies for a given conversation history. Supports multimodal inputs (text, image, video, file), function calling via tools, and both streaming and non-streaming output modes. Returns the completion object including id, model, choices, and usage. Required: model, messages.
- **Create a Z.ai videos generation** (Create). Create a video generation task in Z.AI using CogVideoX or Vidu models from a text prompt, image URL, or pair of first/last frame images. Supports text-to-video, image-to-video, and first/last-frame-to-video workflows. Returns the created generation task. Returns: model, id, request_id, task_status. Required: model.
- **Get single Z.ai paas async result by ID** (Get). Get the result of an asynchronous request in zai by task id. Returns the async result object with task-type-specific fields — either an AsyncVideoGenerationResponse or AsyncImageGenerationResponse depending on the originating task; consult the zai upstream docs for field-level details. Required: id.
- **Create a Z.ai images generation** (Create). Generate a high-quality image from a text prompt using GLM-Image series models in zai. Returns: url, id, created_at. Required: model, prompt.
- **Create a Z.ai audio transcription** (Create). Transcribe an audio file into text in zai using the GLM-ASR-2512 model, supporting multiple languages and optional real-time streaming output. Returns: id, created, request_id, model, text. Required: model.
- **Create a Z.ai paas tokenizer** (Create). Tokenize text input using a specified model in zai to calculate token counts, suitable for text length evaluation, model input estimation, dialogue context truncation, and cost calculation. Returns: created, id, request_id, usage. Required: model, messages.
- **Create a Z.ai paas layout parsing** (Create). Submit a layout parsing request in zai using the GLM-OCR model to extract text content and layout information from a document image or PDF. Returns: id, created, model, md_results, layout_details, layout_visualization, data_info, usage, request_id, attributes. Required: model, file.
- **Create a Z.ai paas web search** (Create). Perform a web search using the zai LLM-optimized Web Search API, which enhances intent recognition to return results better suited for large language model processing, including webpage titles, URLs, summaries, site names, and favicons. Returns: id, created, search_result.
- **Create a Z.ai paas reader** (Create). Create a web reader request in zai to read and parse the content of a specified URL. Returns: id, created, request_id, model, reader_result, url, content. Required: url.
- **Create a Z.ai paas file** (Create). Upload an auxiliary file (such as a glossary or terminology list) to zai to enhance translation accuracy and consistency. Returns the uploaded file record including id, object, bytes, filename, purpose, and created_at. Required: purpose, file. File size limit is 100 MB; accepted formats are pdf, doc, xlsx, ppt, txt, jpg, and png.
- **Create a Z.ai agent** (Create). Create a zai agent task for one of three agent types: General Translation (multilingual text translation with auto-detection and glossary support), Popular Special Effects Videos (AI video generation from an image and prompt via a template), or GLM Slide/Poster (slide or poster generation from natural language instructions). Returns: id, agent_id, choices, usage, status, async_id.
- **Create a Z.ai agents async result** (Create). Query the result of an asynchronous zai agent request. Returns: status, agent_id, async_id, choices, id, usage. Required: async_id.
- **Create a Z.ai agents conversation** (Create). Query conversation history for a zai slides_glm_agent (the only supported agent type). Returns id, agent_id, choices (with messages containing role and content, and finish_reason), and usage token statistics for synchronous responses, or status and async_id for asynchronous responses.
- **Create a Z.ai async images generation** (Create). Create an async image generation job in zai using the GLM-Image model. Returns: model, id, request_id, task_status. Required: model, prompt.