BRIDGE1Use Claude Desktop with a local LLM on Mac
Use Claude Desktop with a local LLM on Mac through Third-Party Inference. Set up the gateway, test the connection and review Code tab limits.
Claude Desktop is tested end to end in real sessions. This does not guarantee every model behaves the same way.
Before connecting Claude Desktop
- For Claude Desktop, start with an Apple Silicon Mac: 24 GB minimum, 32 GB or more recommended. Check memory fit.
- Bridge1 lists Ornith 1.5 35B-A3B (OptiQ 4-bit · default for coding) as its default for coding. Check the model's memory and context before choosing it for Claude Desktop.
- Install Bridge1 through invite-only macOS early access, choose a model and open the Gateway panel for Claude Desktop.
Claude Desktop API and model names
Claude Desktop connects to the Anthropic-format Messages route, /v1/messages, through its Third-Party Inference settings.
Claude Desktop and Claude Code require model names beginning with claude-. They show the same models listed under plain ids in OpenCode, Pi.dev, Codex and Hermes Agent. Keep claude-route1 to follow your Bridge1 selection without changing the client config.
How Route1 chooses a model →Claude Desktop has no Bridge1 launcher; use the manual setup below.
Set up Claude Desktop
Download config.json from Bridge1's Gateway panel, then import it into Claude Desktop.
- In Bridge1’s Gateway panel, select Claude Desktop and choose Download config.json.
- Open Help → Troubleshooting → Enable Developer Mode.
- Open Developer → Configure Third-Party Inference.
- In the configurations switcher, choose New → Import and choose the downloaded file, select Bridge1, then restart Claude Desktop.
- The file contains your gateway key; keep it private and delete it after importing.
First connection check
Use Test connection in Claude Desktop before you start a chat.
Check that it works
- Check the selected Claude Desktop configuration is Bridge1 and its Claude-prefixed model matches the Gateway panel. Restart after importing, then use Test connection; a pass is a successful connection test.
- Start a chat and ask an ordinary question. Look for a relevant local-model reply; a connection test alone does not establish that the model can answer.
- A disposable-project task check is not part of Claude Desktop’s published coverage here. Do not treat a model list, connection test or echoed reply as proof of a completed task.
Check chat separately from the Code tab: the Code tab has the wait and prompt-reread limits below.
If something is off
| Symptom | What it usually means | What to check | More help |
|---|---|---|---|
| Bridge1 is not the selected configuration | The imported configuration still needs to be selected. | Select Bridge1 in the configurations switcher, then restart. | Desktop setup |
| Third-Party Inference controls are missing | The documented import path is not available in your current setup. | Check the Developer Mode steps above; version and plan availability are not recorded. | Desktop coverage |
| Import changed but the chat did not | A restart is part of the manual setup. | Select the imported configuration and restart Claude Desktop. | Claude Code setup |
| Code tab stops while waiting | The documented Code-tab wait limit may apply. | Review the long-prompt and reread conditions below. | Code-tab limits |
Versions
Claude Desktop uses manual setup; Bridge1 0.7.7 has no launcher for it.
Claude Desktop: limits to check
- The Code tab gives up after about 13 minutes without visible output, for example while reading a very long prompt.
- In the Code tab, the first step of a new prompt reread the whole chat (about 2.5 minutes at 53K tokens on Qwen3.6-35B); later steps reused it.
Other client setup guides
Request Bridge1 early accessClaude Desktop setup questions
Where do I connect Claude Desktop to Bridge1?
Download config.json from the Gateway panel. Enable Developer Mode, then open Developer → Configure Third-Party Inference. In the configurations switcher, choose New → Import, choose the downloaded file, select Bridge1 and restart Claude Desktop.
What should I check before a Claude Desktop chat?
Select the Bridge1 configuration, restart Claude Desktop and use Test connection before starting a chat. The optional X-Bridge1-Client: claude-desktop header lets the Live Log identify it.
Which Claude Desktop limits apply to the Code tab?
The Code tab gives up after about 13 minutes without visible output. The first step of a new prompt reread the whole chat (about 2.5 minutes at 53K tokens on Qwen3.6-35B); later steps reused it.