← Volver a las guías

Claude Desktop to Custom API: 59API Quick Start

Why this setup matters

If you want Claude Desktop to work with a custom API endpoint, the key thing to know is this: the desktop app is not a generic model picker. In most cases, you do not swap the backend by pasting in a URL and calling it done. The reliable approach is to connect Claude Desktop to a local bridge or MCP server, then have that bridge forward requests to your endpoint.

That is where 59API fits nicely. It gives you pay-as-you-go access to Claude models and GPT models through https://api.59api.com, with native official-quality models, no downgrade, and pricing that is usually much cheaper than building your own relay. If you are testing, shipping a side project, or trying to keep per-request costs under control, it is a very practical backend.

Step 1: Get your 59API access ready

Create a 59API account, copy your API key, and keep the base URL handy: https://api.59api.com. The relay is OpenAI-compatible, which makes it easy to reuse across tools and bridges. That compatibility matters because once your bridge knows the base URL and key, you can point different clients at the same infrastructure without rewriting your app each time.

• Base URL: https://api.59api.com
• Auth: your 59API API key
• Models: Claude Opus, Sonnet, Haiku, Fable, plus GPT models

For busy developers, the biggest win is speed: one account, one relay, and pay only for what you use.

Step 2: Add a small bridge between Claude Desktop and the API

Claude Desktop works best with an adapter layer. If you already have an internal service, make that service your custom endpoint and have it call 59API upstream. If you do not, use a small local MCP server or proxy that exposes the action you want and forwards the request to 59API.

Think of the bridge as a thin translation layer. Claude Desktop talks to the bridge locally. The bridge sends the request to 59API. 59API returns the model response. This pattern keeps your desktop setup clean and lets you change providers later if you ever need to.

• Local side: Claude Desktop
• Middle layer: MCP server or lightweight proxy
• Upstream: https://api.59api.com

If your bridge already supports OpenAI-style configuration, you usually only need to set the base URL and API key. For many tools that means two values: a base URL field pointing at https://api.59api.com and an auth field containing your key.

Step 3: Register the bridge in Claude Desktop

Open Claude Desktop and edit its MCP configuration file. On macOS, the file is usually in ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows, check %APPDATA%\Claude\claude_desktop_config.json. Add your local bridge as an MCP server, then restart Claude Desktop so it reloads the config.

The exact bridge command depends on your setup, but the pattern is always the same: Claude Desktop launches or connects to the local server, and that server knows how to reach 59API. If you are using an existing adapter, point it at https://api.59api.com and store the key in an environment variable rather than hard-coding it.

Step 4: Test with a simple prompt

After restart, open Claude Desktop and send a test request that forces the bridge to do something obvious. For example, ask it to summarize a short file, transform a JSON payload, or call a tool exposed by your server. If the bridge is wired correctly, you should see the response come back from the custom path rather than from a direct local-only action.

When you test, watch for three common issues:

• Wrong base URL: make sure it is https://api.59api.com and not a typo.
• Missing key: confirm your API key is loaded in the bridge process.
• Stale config: restart Claude Desktop after every config change.

Step 5: Choose the right model for the job

One reason 59API is attractive is that you are not forced into a single expensive model. Use Opus for the hardest reasoning tasks, Sonnet for most day-to-day coding work, and Haiku when latency matters more than depth. If Fable is the right fit for your workflow, it is there too. Because the relay uses official-quality native models, you are not taking a quality hit just to save money.

That makes the setup especially useful for teams that prototype in Claude Desktop, then move the same endpoint into scripts, internal tools, or production clients later. The same relay also works with Claude Code, Codex, and any OpenAI SDK, so you are not locked into one app.

Bottom line

If you need Claude Desktop to work with a custom API endpoint, the fastest path is a local bridge plus an OpenAI-compatible backend. 59API is a strong choice because it is cheap, pay-as-you-go, and already built for this kind of integration. If you want to try it, sign up for 59API, point your bridge at https://api.59api.com, and you can be testing in minutes instead of building your own relay from scratch.

And if you are sharing access with teammates, the referral rebate is a nice extra way to lower costs while you scale usage.

Precios · Docs