How to use Zed's AI agent with a self-hosted model on a cloud GPU

Rent a GPU on Deploy, then open a terminal in JupyterLab or over SSH. Serve an open-weight model on that machine, and set Zed’s language_models Ollama or OpenAI-compatible block to that endpoint. Open the Agent Panel the way the agent-panel page prints, type a message, and press enter. Copy the project the agent edits before you destroy the instance, because destroying the instance deletes its disk and stops the server.

An agent can rent the GPU and serve the model the editor calls. The agent pages are GPU compute for AI agents and the agent integration guide. Those pages are how an agent discovers a GPU and runs work on it. This page is the Zed setting. An editor whose Ollama config sets baseURL to http://localhost:11434/v1 is How to use Kilo Code with a self-hosted model on a cloud GPU. A coding assistant whose Docker command publishes port 8080 and passes --model and --chat-model is How to run Tabby self-hosted coding assistant on a cloud GPU. A terminal client whose OpenAI-compatible page sets OPENAI_API_BASE is How to use Aider with a self-hosted model on a cloud GPU. An editor client whose install page is the VS Code extension is How to use Cline with a self-hosted model on a cloud GPU. An editor whose config.yaml sets an apiBase for chat, autocomplete, and embeddings is How to use Continue with a self-hosted model on a cloud GPU. An editor whose Ollama provider defaults to http://localhost:11434 is How to use Roo Code with a self-hosted model on a cloud GPU. A coding agent whose providers page sets package to the OpenAI-compatible provider is How to use OpenCode with a self-hosted model on a cloud GPU. A coding-agent server whose card prints --tool-call-parser qwen3_coder is How to serve Qwen3-Coder for coding agents on a cloud GPU. A platform whose single-container guide sets CHAT_BASE_URL to http://host.docker.internal:11434/v1 for a local OpenAI-compatible server is How to run AutoGPT with a self-hosted model on a cloud GPU. An agent whose Docker command publishes port 5080, and whose Ollama provider sets api_base to http://host.docker.internal:11434, is How to run Agent Zero with a self-hosted model on a cloud GPU. An agent whose Ollama example sets base_url to http://localhost:11434/v1 is How to run OpenManus with a self-hosted model on a cloud GPU. A chat app whose Docker page runs docker compose up -d and whose Ollama page sets baseURL to http://localhost:11434/v1/ is How to run LibreChat agents with a self-hosted model on a cloud GPU. A chat app whose README runs docker compose up -d and whose environment page sets OLLAMA_PROXY_URL to http://127.0.0.1:11434 is How to run LobeChat agents with a self-hosted model on a cloud GPU. A terminal agent whose install page runs curl -fsSL https://www.openinterpreter.com/install | sh, and whose models page sets the Ollama base URL to http://localhost:11434/v1, is How to run Open Interpreter with a self-hosted model on a cloud GPU. A workspace app whose Docker quickstart publishes port 3001, and whose Ollama page sets the URL to http://127.0.0.1:11434, is How to run AnythingLLM agents with a self-hosted model on a cloud GPU. A repository agent whose local-models section sets api_base to http://localhost:11434 is How to run SWE-agent with a self-hosted model on a cloud GPU. A workflow editor whose Docker install publishes port 5678, and whose Ollama credentials default the Base URL to http://localhost:11434, is How to run n8n AI agents with a self-hosted model on a cloud GPU. A CLI whose providers page sets OLLAMA_HOST to http://localhost:11434 is How to run the goose agent with a self-hosted model on a cloud GPU. A canvas whose self-host page opens http://localhost/install is How to run Dify with a self-hosted model on a cloud GPU. A canvas whose ChatOllama page names a base URL is How to run Flowise with a self-hosted model on a cloud GPU. A flow editor whose Ollama component defaults the base URL to http://127.0.0.1:11434 is How to run Langflow with a self-hosted model on a cloud GPU. A CLI whose connect help sets --base-url to http://localhost:8000/v1 for openai-compatible is How to run Letta agents with a self-hosted model on a cloud GPU. A browser chat whose serve command listens on port 8080 is How to run Open WebUI on a cloud GPU. A Python agent library whose CodeAgent can take TransformersModel or an OpenAI-compatible server is How to run smolagents with a self-hosted model on a cloud GPU. A crew whose LLM class takes base_url for Ollama or an OpenAI-compatible server is How to run CrewAI with a self-hosted model on a cloud GPU. A graph runtime whose overview example does not name a model is How to run LangGraph agents with a self-hosted model on a cloud GPU. An AgentChat library whose Ollama example sets model to llama3.2 is How to run AutoGen agents with a self-hosted model on a cloud GPU. A typed agent whose vLLM example sets base_url to http://localhost:8000/v1 is How to run Pydantic AI agents with a self-hosted model on a cloud GPU. A framework whose README sets Ollama and HuggingFaceEmbedding is How to build LlamaIndex agents with a self-hosted model on a cloud GPU. A browser agent whose models page sets base_url to http://localhost:11434/v1 is How to run browser-use with a self-hosted model on a cloud GPU. A pipeline whose OllamaChatGenerator sets url to http://localhost:11434 is How to run Haystack agents with a self-hosted model on a cloud GPU. A program whose language-models page sets dspy.LM to ollama_chat/llama3.2:1b is How to run DSPy with a self-hosted model on a cloud GPU. An agent whose Ollama example sets id to llama3.1:8b is How to run Agno agents with a self-hosted model on a cloud GPU. A TypeScript agent whose local-models section sets url on the model object is How to run Mastra agents with a self-hosted model on a cloud GPU. A ChatAgent whose vLLM example sets ModelPlatformType.VLLM is How to run CAMEL-AI agents with a self-hosted model on a cloud GPU. A local agent UI whose Docker command publishes port 3000 is How to run OpenHands with a self-hosted model on a cloud GPU. The README install is pip install --upgrade metagpt: How to run MetaGPT with a self-hosted model on a cloud GPU. A general vLLM install is How to serve an OpenAI-compatible API with vLLM on a cloud GPU. A general SGLang install is How to run SGLang on a cloud GPU. Ollama’s install and port 11434 are How to run Ollama on a cloud GPU. CloudGPU’s hosted gateway is a different base URL: How do I use an OpenAI-compatible API for open models?. The Cursor fields on How do I use CloudGPU as a custom base URL in Cursor? are for that hosted URL, not for the localhost URL of a server you start. How published weight sizes are read: How much VRAM do I need to run an LLM?. The tunnel is How to use SSH port forwarding on a cloud GPU. If the process runs out of memory: How to fix CUDA out of memory on a cloud GPU. If the server raises no kernel image on an RTX 5090: How to fix CUDA no kernel image on an RTX 5090.

The README says Zed source code is licensed primarily under GPL-3.0-or-later, with Apache-2.0 components where marked.

The settings below are copied from the Zed docs named in each step, fetched 2026-11-14. They were not run on a CloudGPU instance for this page. Anything this page adds around them is marked. The official repository is zed-industries/zed. The providers page and the API-access page name hosted providers. This page does not copy those names. It copies the local-model page’s Ollama and LM Studio sections, and the API-access page’s OpenAI-compatible section.

What the Zed pages print about memory

The local-model page and the API-access page, fetched 2026-11-14, print no GPU GB figure. The Ollama block sets max_tokens to 32768 on one example model, and it says Zed uses 4096 tokens as the default num_ctx. A later block sets context_window to 8192. Those are token counts. This page does not treat them as vram_gb. The API-access example sets max_tokens to 128000. That is a context-window field. This page does not treat it as vram_gb. This page does not mark a catalog row as enough.

The llama.cpp section on the local-model page prints model ids this page does not copy. The keys in that JSON are api_url, auto_discover, and available_models. This page follows the Ollama section, whose example model name is qwen2.5-coder, and the pull line ollama pull mistral.

GET /v1/public/compute-capabilities does not list GPUs. It returns feature flags. The memory column is vram_gb from GET /v1/public/gpu-catalog (generated_at 2026-11-14T14:09:20.887Z). The catalog response also includes an availability field. This page does not copy it. Every catalog row below has min_gpu_count 1. Rows are sorted by vram_gb, then by name. Which listing you can start is on Deploy.

The catalog template list had no Zed template. It did have Ollama templates (jupyterlab-ollama, ollama-qwen3, ollama). Those templates are a different install. This page does not treat them as Zed.

Catalog name vram_gb
RTX 2080 8
RTX 3080 10
RTX 2080 Ti 11
RTX 3080 Ti 12
V100 16G 16
A30 24
RTX 3090 24
RTX 4090 24
RTX 4090D 24
RTX 4080 SUPER 32G 32
RTX 5090 32
RTX 5090D 32
V100 32G 32
A100 40G 40
A800 40G 40
A40 48
RTX 4090 48G 48
RTX 4090D 48G 48
RTX A6000 48
Ascend 910B 64
NVIDIA RTX PRO 5000 72G 72
RTX PRO 5000 72G 72
A100 80G 80
A800 80G 80
H800 80
RTX 6000D 83
H20 96G 96
RTX PRO 6000 96G 96
RTX PRO 6000 Max-Q 96
H20-3e 141G 141

Ollama and LM Studio

The local-model page’s Ollama steps, copied:

  1. Download and install Ollama from ollama.com/download.
  2. Pull a model: ollama pull mistral
  3. Make sure the Ollama server is running. On Linux or from a shell, run ollama serve
  4. In Zed, select an Ollama model from the model dropdown.

The page says Zed discovers models Ollama has pulled. To list models yourself, it prints this settings block:

{
  "language_models": {
    "ollama": {
      "api_url": "http://localhost:11434",
      "auto_discover": false,
      "available_models": [
        {
          "name": "qwen2.5-coder",
          "display_name": "qwen 2.5 coder",
          "max_tokens": 32768,
          "supports_tools": true,
          "supports_thinking": true,
          "supports_images": true
        }
      ]
    }
  }
}

It says Zed sends num_ctx and uses 4096 tokens by default. The context-window block, copied:

{
  "language_models": {
    "ollama": {
      "context_window": 8192
    }
  }
}

It says a per-model length is max_tokens in available_models. If the server requires a key, it says to enter the key in the provider UI or set OLLAMA_API_KEY. For a remote server, it says to set the API URL to the remote endpoint and provide an API key. It does not print an ssh command. How to use SSH port forwarding on a cloud GPU is how a program on your computer reaches a port on the instance. This page does not add an ssh command. After a forward that lands on port 11434 of your computer, the api_url the page already prints is http://localhost:11434. That URL has no /v1 suffix. This page does not add one.

The LM Studio steps, copied:

  1. Download and install LM Studio.
  2. Download at least one model in LM Studio, or use the LM Studio CLI: lms get qwen2.5-coder-7b
  3. Start the LM Studio API server: lms server start
  4. In Zed, select an LM Studio model from the model dropdown.

The page does not print a port for lms server start. It does not print an api_url for LM Studio. This page does not invent either. If the server requires a key, the page says to enter it in the provider UI or set LMSTUDIO_API_KEY.

The local-model page’s OpenAI-compatible section is one sentence: use OpenAI-compatible endpoints for local or self-hosted servers that expose an OpenAI-compatible API. That section does not print api_url or available_models. Those keys are on the API-access page, below.

The Zed pages this guide read do not print hf download or huggingface-cli. The pull command they print is ollama pull mistral, and the LM Studio command is lms get qwen2.5-coder-7b. How to download a Hugging Face model is a different download. This page does not paste it into Zed’s settings.

OpenAI-compatible settings

The providers page says OpenAI-compatible provider setup has moved to the API-access page. It also lists hosted, subscription, and gateway paths. This page does not copy those names. The local-model path is the one above.

The API-access page’s OpenAI-compatible section says to use it when you have a custom base URL, model ID, and API key. It says to open Agent Settings with agent: open settings, look for Add Provider in the LLM Providers section, and fill in the provider name, API URL, model ID, and context window. The settings file it prints, copied:

{
  "language_models": {
    "openai_compatible": {
      "my-provider": {
        "api_url": "https://example.com/v1",
        "available_models": [
          {
            "name": "my-model",
            "display_name": "My Model",
            "max_tokens": 128000
          }
        ]
      }
    }
  }
}

my-provider, https://example.com/v1, and my-model are the placeholders that page prints. This page does not replace them with a host or a model id the example does not print. A vLLM server’s listen address is a different document: How to serve an OpenAI-compatible API with vLLM on a cloud GPU. This page does not merge that port into the example.

The page says OpenAI-compatible models inherit tools true, images false, parallel_tool_calls false, prompt_cache_key false, chat_completions true, interleaved_reasoning false, and max_tokens_parameter false. It says to set capabilities.chat_completions to false when a model only works with a different endpoint. A later example on that page uses a model id this page does not copy. The effort values that paragraph prints are none, minimal, low, medium, high, xhigh, and max.

It says to enter the API key in the provider settings UI or set the generated environment variable, and not to put API keys in settings.json. The environment-variable rule it prints is: a provider id my-gateway uses MY_GATEWAY_API_KEY. It says keys saved through Zed are stored in the system keychain. It says that in SSH, dev container, and other remote projects, those keys are read from the local Zed process, not from the remote machine. That sentence is not an ssh port-forward command. The page does not print one.

Agent Panel

The agent-panel page says to open the panel with agent: new thread from the Command Palette, or click the sparkles icon in the status bar. It says that with a provider configured, you type in the message editor and press enter. It does not print a sample prompt. This page does not invent one. It says agent: open active thread as markdown opens the thread as Markdown, and that the context menu includes Open Thread as Markdown. The page’s image-context paragraph names hosted model families. This page does not copy those names.

This page did not open the panel.

When it fails

No tools. The agent-panel page says tool calling has to be supported by the model, and that a model without that support shows a “No tools” label. The Ollama example sets supports_tools to true for qwen2.5-coder. This page did not check that tag’s tool support on a server.

CUDA out of memory. The Zed pages this guide read print no GPU GB figure. If the model server raises CUDA out of memory, use how to fix CUDA out of memory on a cloud GPU. This page does not pick a smaller model id for you.

“CUDA error: no kernel image is available for execution on the device” on an RTX 5090. The Zed pages this guide read do not pin a PyTorch build and do not name compute capability 12.0. The RTX 5090 guide says that card is compute capability 12.0 (sm_120). This page did not re-test a model server on an RTX 5090. Catalog RTX 5090 and RTX 5090D are 32. That is the catalog number only.

The local-model page does not print the OpenAI-compatible JSON. That JSON is on the API-access page. This page does not copy the local-model page’s one-sentence section into the API-access JSON, and it does not invent fields the local-model section does not print.

What to copy off, then destroy

The agent edits files in the project you opened. The agent-panel page says Open Thread as Markdown writes the thread to a tab. Copy the project, and that Markdown if you saved it, before you destroy the instance. The model server stops with the instance. Save your work before you destroy a cloud GPU is the checklist. The terms say data on local instance storage is permanently deleted when the instance is destroyed.

The billing page states the billing rule in these words:

Per-minute options have a 1-minute minimum; unused prepaid minutes are refunded when you destroy the instance. Some configurations use whole-hour billing with a 1-hour minimum; unused time within a charged hour is not refunded on voluntary stop. The billing period, minimum charge and initial prepayment are shown before confirmation. No subscription.

Amounts are not copied here.

Frequently asked questions

What memory figure do the Zed pages print?

The local-model page and the API-access page print no GPU GB figure. The Ollama example sets max_tokens to 32768. The page says the default num_ctx is 4096 tokens, and one block sets context_window to 8192. The OpenAI-compatible example sets max_tokens to 128000. Those are token counts. This page does not treat them as vram_gb and does not mark a catalog row as enough. Which listing you can start is on https://cloudgpu.app/deploy.

Which settings point Zed at a server on the GPU?

The Ollama block sets api_url to http://localhost:11434 and lists available_models. The example name is qwen2.5-coder. The pull line is ollama pull mistral, and the serve line is ollama serve. The OpenAI-compatible block is language_models.openai_compatible. The example sets api_url to https://example.com/v1 and the model name to my-model. The local-model page’s OpenAI-compatible section does not print those keys. Neither page prints an SSH command. This page did not run them.

How does the agent-panel page say to send a task?

Open the panel with agent: new thread, or click the sparkles icon in the status bar. Type in the message editor and press enter. The page does not print a sample prompt. agent: open active thread as markdown opens the thread as Markdown. This page did not open the panel.

What license sentence does the README print?

The README says Zed source code is licensed primarily under GPL-3.0-or-later, with Apache-2.0 components where marked. LICENSE-GPL begins with GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007. LICENSE-APACHE begins with Copyright 2022 - 2025 Zed Industries, Inc. The GitHub license field on 2026-11-14 was Other, SPDX NOASSERTION. Repository HEAD was 99c890d39d336052d9dae34c7914b55ee82abc5f, committed 2026-11-14T13:43:29Z. That license text is not a CloudGPU rate.

Sources (read 2026-11-14, Asia/Shanghai)

  • Zed local-model page: https://zed.dev/docs/ai/use-a-local-model . Ollama api_url http://localhost:11434, available_models name qwen2.5-coder, max_tokens 32768, default num_ctx 4096, context_window 8192, ollama pull mistral, ollama serve, OLLAMA_API_KEY. LM Studio lms get qwen2.5-coder-7b, lms server start, no port printed. The llama.cpp model ids are not copied. The OpenAI-compatible section does not print api_url or available_models. No GPU GB figure. No SSH command
  • Zed API-access page: https://zed.dev/docs/ai/use-api-access . language_models.openai_compatible, example api_url https://example.com/v1, model name my-model, max_tokens 128000. Inherited capability defaults. Effort values none, minimal, low, medium, high, xhigh, max. Provider id my-gateway uses MY_GATEWAY_API_KEY. Keys stay in the local keychain. A later example’s model id is not copied. Hosted provider names are not copied
  • Zed providers page: https://zed.dev/docs/ai/llm-providers . OpenAI-compatible setup has moved to the API-access page. Hosted, subscription, and gateway names are not copied
  • Zed agent-panel page: https://zed.dev/docs/ai/agent-panel . agent: new thread. Sparkles icon. Type and press enter. No sample prompt. agent: open active thread as markdown. Image-context model names are not copied
  • zed-industries/zed README at commit 99c890d39d336052d9dae34c7914b55ee82abc5f, committed 2026-11-14T13:43:29Z. “licensed primarily under GPL-3.0-or-later, with Apache-2.0 components where marked.” LICENSE-GPL begins with GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007. LICENSE-APACHE begins with Copyright 2022 - 2025 Zed Industries, Inc. GitHub license field Other, SPDX NOASSERTION. The GPL preamble is not copied
  • The pages above do not print hf download or an ssh command
  • Public GPU memory: https://cloudgpu.app/v1/public/gpu-catalog (generated_at 2026-11-14T14:09:20.887Z). GET /v1/public/compute-capabilities returned feature flags and no GPU list. The catalog had no Zed template. It listed Ollama template ids jupyterlab-ollama, ollama-qwen3, and ollama

Cards you can start are on the Deploy page. Amounts are not copied here.

Last updated: 2026-11-14 (Asia/Shanghai).