Continue with your agent
Bring this model and your idea into your agent’s chat.
Your starting point
Opening your selected model…
01 / Connect
Connect Blendstra to your agent.
Choose your agent below. Add Blendstra in its connection settings, then approve your workspace and permissions.
https://blendstra.com/mcp
ChatGPT Work setup
Use the personal-plugin setup in ChatGPT Work: add this public MCP server in Developer mode, install your personal plugin, then mention it in your work chat. Follow the client’s sign-in flow and approve the intended Blendstra workspace. Available setup controls depend on your account and workspace policy.
https://blendstra.com/mcp
Official ChatGPT Work instructions · Checked 2026-09-07.
Your account
Connected agents
Connection troubleshooting
Connection health
Connection diagnostics
Reading current endpoint metadata…
These checks do not run Blender, spend an allowance or verify your external client.
Previous saved requests
Requests saved through the earlier workflow are still available here.
Workspace access and permissions
- Choose one workspaceA connection acts within the workspace you approve and your current project roles.
- Connect onceKeep Stay connected enabled for automatic renewal. Your agent can work with your models when you ask. Sharing is optional, and you can adjust individual permissions in Advanced details.
- Revoke when you are doneRemove a saved connection below to block future authorized requests. Copies already downloaded remain on their recipients’ devices.
Help with connections and model jobs
| What you see | What to do |
|---|---|
| Connection keeps expiring | Open Connected agents below and choose Keep connected. If automatic renewal was not enabled, reconnect from your agent once and leave Stay connected enabled. |
| Permission or workspace denied | Choose the workspace that contains your project. If it belongs to someone else, ask its owner for access. |
| Agent cannot connect | Check the connection address and complete sign-in from your agent. Connection diagnostics can help identify a service problem. |
| A model request reaches a limit | Keep the saved project and ask your agent to split the change into smaller steps. If the limit persists, contact the workspace owner. |
| A saved-version conflict | Open the latest saved version and compare changes before trying your edit again. |
| Interrupted request | Ask your agent to check whether the model was saved before starting again. Your earlier versions remain available. |
Developer resources and compatibility
MCP and REST operate on the same projects. Blendstra asks for no model-provider API key and runs no hosted inference. Astra, Fable or another model is chosen in your external client; model names do not establish transport compatibility.
Setup and sign-in options depend on your agent and account. A responding connection does not confirm that a model request has completed.