Symbioza
Menu

Symbioza / Integrations

Run GPU jobs from your assistant.

Connect ChatGPT, Claude or Cursor. Ask for a free estimate, set a spending limit and get the output files back.

ChatGPT

Available in ChatGPT

Symbioza is in ChatGPT’s plugin directory. Open the listing, or add Symbioza as a custom MCP app if your plan and workspace allow it.

ChatGPT and the OpenAI logo are trademarks of OpenAI. Symbioza is not endorsed by OpenAI.

Available in ChatGPT · Or connect with custom MCP

  1. Check that you can add custom apps

    You can use Symbioza today if your ChatGPT plan and workspace allow custom MCP apps. Some accounts must turn on Developer mode first, and on a company or school workspace an admin may need to allow custom apps.

  2. Add Symbioza

    Open Plugins and choose Add → Add custom MCP server. Name it Symbioza, paste the server URL, keep Authentication on OAuth, confirm the custom-server notice and choose Create as a plugin.

    Didn’t open? chatgpt.com/plugins (opens in a new tab)

  3. Sign in, then use it in a chat

    ChatGPT opens the Symbioza sign-in. Use the account you use here, then select Symbioza in your chat.

Check it works

Ready when: you ask 'Use Symbioza to list my jobs.' and get a list — an empty one counts.

This check is free and starts no work.

Troubleshooting

No “Add custom MCP server”? Your plan or workspace does not allow custom MCP apps yet: turn on Developer mode if your account offers it, or ask your workspace admin.

The plugin will not save? The URL is exactly https://symbioza.dev/mcp, with no trailing path.

“Unauthorized”? Reconnect and complete the sign-in with the account you use here.

Already added? Select Symbioza in your chat. No need to add it again.

ChatGPT setup help · More troubleshooting

Claude

App or Claude Code

Add Symbioza as a custom connector in Claude on the web or desktop, then sign in.

Install the Symbioza plugin from your terminal, then sign in with /mcp.

Claude app (web and desktop)

  1. Add the connector

    Open Customize → Connectors, choose Add connector → Add custom connector, name it Symbioza and paste the server URL.

    Didn’t open? claude.ai/customize/connectors (opens in a new tab)

  2. Sign in to Symbioza

    Connect it and complete the Symbioza sign-in that opens. Using a team account? Your workspace owner may need to add it first.

  3. Turn it on in a conversation

    In a new chat, open the connector menu and turn on Symbioza.

Claude Code

  1. Install the plugin

    Run both commands in your terminal.

    The install ends with Successfully installed plugin: symbioza@symbioza.

  2. Start a new session

    Start a new Claude Code session so it loads the plugin.

  3. Sign in to Symbioza

    Run /mcp, select plugin:symbioza:symbioza and choose Authenticate. Your browser opens to sign in.

Check it works

Ready when: you ask 'Use Symbioza to list my jobs.' and get a list — an empty one counts.

Have files to process? Ask Claude to check that Symbioza can access them. A chat attachment alone isn’t enough.

Troubleshooting

“Unauthorized”? Reconnect the connector and complete the sign-in with the account you use here.

Symbioza missing from the chat? Turn it on in the connector menu of a new chat.

Claude’s connector guide (opens in a new tab) · More troubleshooting

Check it works

Ready when: you ask 'Use Symbioza to list my jobs.' and get a list — an empty one counts.

Your code runs remotely. Ask Claude to check that the job can access its code and input files.

Troubleshooting

Every tool listed twice? The plugin and claude mcp add each add the connector. Keep one: claude mcp remove symbioza.

No plugin:symbioza:symbioza in /mcp? Start a new Claude Code session after the install.

Plugin on GitHub (opens in a new tab) · Claude Code setup help

Cursor

MCP config · editor

Add Symbioza to your Cursor MCP config, then sign in from Cursor’s Customize panel.

  1. Add the connection

    Add Symbioza to ~/.cursor/mcp.json for every project, or to .cursor/mcp.json in one. Already have servers there? Add the symbioza entry inside your existing mcpServers block.

    Or install the plugin: tools plus the job-preparation skill

    The plugin adds the run-on-symbioza skill. Copy it into Cursor’s local plugins folder, then restart Cursor or run Developer: Reload Window.

    Choose the plugin or the config above, not both, or your agent sees every tool twice. A team policy may block local plugins.

  2. Sign in to Symbioza

    Open Customize in the sidebar, enable symbioza under MCPs and finish the sign-in in your browser. There is no key to paste.

Check it works

Ready when: you ask 'Use Symbioza to list my jobs.' and get a list — an empty one counts.

Then ask for a free estimate. It needs no spending limit and submits nothing.

Troubleshooting

Every tool listed twice? Keep either the config entry or the plugin, not both.

Symbioza not under MCPs? Check that the JSON is valid, then restart Cursor or run Developer: Reload Window.

Cursor guide (opens in a new tab) · More troubleshooting

  • Free estimates
  • Runs use your Symbioza credit
  • You set the spending limit

See how it works Another MCP client?

HOW IT WORKS

Ask. Review. Run.

Describe the work and share its inputs. Your assistant helps with the job; Symbioza runs it on remote compute and collects the output files.

  • You bringCode, inputs & a spending limit
  • Your assistant helps withA supported image & command
  • Symbioza handlesThe run, progress & output files
MODEL EVALUATIONILLUSTRATIVE EXAMPLE
Explore the job workflow

Ask your assistant

“Help me run this evaluation on a GPU. Check what’s missing and estimate it. Don’t submit yet.”
Supported imageCommand + inputsBudget

The estimate is free. Review its price and runtime limit before submitting.

After you choose to submit

Your job runs on remote compute.

QueuedPreparingExecuting

Use the execution ID to check progress through your assistant or your Symbioza account. The run uses your prepaid balance and job budget.

Outputs from your workload

Get your files back.

metrics.jsonEXAMPLE FILE
predictions.csvEXAMPLE FILE

Download the files your job produced. Check the delivery status and whether the charge is final.

Explore the steps. This example submits no work.

Bring your own workload

More than your laptop can handle.

Bring your code and data. Your assistant helps prepare them for the run.

01 / TRAIN

Fine-tune a model

Run your training script on a GPU. Bring back checkpoints and logs.

CHECKPOINTS + TRAINING LOGS

02 / EVALUATE

Evaluate a model

Test a model against your dataset. Collect metrics and predictions.

METRICS + PREDICTIONS

03 / PROCESS

Process a dataset

Upscale images, run inference or process a batch. Download the output files.

PROCESSED DATA + SUMMARY

Explore example jobs

MCP & agent documentation

Connect your own agent.

Use the MCP endpoint with your Symbioza account. The guide covers sign-in, job setup, spending limits and downloads.

Read the MCP & agent guide

Using another MCP client?

It must support a remote HTTP MCP server and OAuth sign-in. Add the URL, authenticate and verify the nine tools. Client support varies; check setup before submitting work.

Manual setup guide ↗

Authentication9 toolsJob requirementsBudgets & outputs

Before your first run

Before you run.

More details in pricing and the agent guide.

Can I just describe what I want?

Start with the task, your files and a spending limit. Your assistant can help prepare the command and a supported container image, and check that the remote job can reach its inputs.

What is free, and when do I pay?

Connecting, estimates, status checks and listing your own jobs are free. Submitted jobs use prepaid credit. A failed or cancelled job can still incur charges for costs already incurred. Review the billing details before your first run.

Where do my files and data go?

Your job runs on third-party compute. Put output files in /workspace/artifacts/ for collection. Retrieve available files and check their delivery status. Keep credentials out of job specifications. See our privacy information.

How long can a job run?

Review the runtime limit in the estimate. It depends on your budget, your requested limit and the compute available. A machine’s lifetime can also constrain the run. Recovery depends on job policy, remaining budget and available compute.

Which connection should I choose?

Choose the assistant you already use. ChatGPT and Claude web/Desktop connect through MCP. The Claude Code plugin adds job-preparation guidance for working from your repository. In Cursor, connect directly for the tools, or install the plugin for the tools plus that guidance. All use your Symbioza account; connector availability can depend on your plan and workspace settings.

Start with a free estimate.

Choose your assistant and try a task.

Choose your assistant