Cursor alternatives in 2026: eight tools that work with your own API key

People look for Cursor alternatives for different reasons: they want to work in a terminal instead of an editor, stay in JetBrains or Zed, or choose the model themselves and pay per token rather than per subscription. Below are eight tools: where each one runs (per their official docs as of October 2026) and how to connect it to a single Zukko key with the right API address.

When Cursor isn't a fit

Cursor can work with your own key too: in the model settings you enable OpenAI API Key and Override OpenAI Base URL with https://api.zukko.pro/v1. But that requires a paid Cursor Pro plan, and popular models need special names — details in the Cursor guide↗. If that's inconvenient, or you simply work in a different environment, here are the options.

Cursor alternatives: summary table

The address rule: clients that append /v1 themselves (Claude Code, Roo Code's Anthropic provider) get the address without /v1. OpenAI-compatible clients and OpenCode get it with /v1. More on choosing the format in the docs↗.

Terminal agents

Claude Code

Anthropic's agent: reads code, edits files, runs commands. Per its docs it is available in the terminal, IDE extensions, a desktop app and the browser; the CLI, VS Code and JetBrains support third-party providers.

export ANTHROPIC_BASE_URL=https://api.zukko.pro
export ANTHROPIC_AUTH_TOKEN=fetch_your_key
export ANTHROPIC_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-sonnet-5
claude

The key goes into ANTHROPIC_AUTH_TOKEN. ANTHROPIC_DEFAULT_HAIKU_MODEL is required: Claude Code sends background requests to the Haiku-slot model (by default claude-haiku-*), which is not in the Zukko catalog, and the gateway answers an unknown model name with 404 model_not_found. For /model opus and /model sonnet, set ANTHROPIC_DEFAULT_OPUS_MODEL and ANTHROPIC_DEFAULT_SONNET_MODEL the same way. Claude Code uses the Anthropic format, so prompt caching is counted for Claude.

Codex

OpenAI's agent that runs locally; the repository is Apache-2.0 licensed. Besides the CLI there is an extension for VS Code, Cursor and Windsurf, and a desktop app. A custom provider goes into ~/.codex/config.toml:

model = "gpt-6.1-sol"
model_provider = "zukko"

[model_providers.zukko]
name = "Zukko"
base_url = "https://api.zukko.pro/v1"
env_key = "ZUKKO_API_KEY"
wire_api = "responses"

Put the key into the ZUKKO_API_KEY environment variable. Codex talks through the Responses API: GPT and Claude models work with it, GLM models don't. Step-by-step Codex setup is in the Zukko docs under IDE Integrations.

OpenCode

An open source agent: a terminal UI, desktop app or IDE extension. It connects through the anthropic provider in opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "anthropic": {
      "options": {
        "baseURL": "https://api.zukko.pro/v1",
        "apiKey": "fetch_your_key"
      },
      "models": {
        "claude-sonnet-5": {},
        "claude-opus-5-5": {}
      }
    }
  }
}

Here the address includes /v1: without it requests go to /messages and get a 404.

Extensions for VS Code and JetBrains

Cline

An agent that reads files, writes code and runs commands with your approval. It installs as an extension in VS Code, Cursor, JetBrains, Windsurf and VSCodium, and also comes as a CLI and a desktop app. Setup: in Cline settings choose API Provider "OpenAI Compatible", enter Base URL https://api.zukko.pro/v1, your fetch_... key and a Model ID such as claude-sonnet-5 or gpt-6.1-sol.

Roo Code

Per its docs, an open source agent for VS Code with file system and terminal access. For Claude models choose the "Anthropic" provider, paste the key into the Anthropic API Key field, enable "Use custom base URL" and enter https://api.zukko.pro. For other models use "OpenAI Compatible" with https://api.zukko.pro/v1. The Roo Code docs warn that the model must support native tool calling.

Kilo Code

Per its docs, an open source platform: extensions for VS Code, JetBrains and VS Code forks, a CLI and a desktop app. Add your own provider in settings on the providers tab via "Custom provider": type "OpenAI Compatible", Base URL https://api.zukko.pro/v1, your key. Kilo fetches the model list from /v1/models, so you can pick IDs from a list.

Continue

Open source extensions for VS Code and JetBrains, plus the cn CLI. Models are described in ~/.continue/config.yaml:

name: Zukko
version: 0.0.1
schema: v1
models:
  - name: Claude Sonnet 5
    provider: openai
    model: claude-sonnet-5
    apiBase: https://api.zukko.pro/v1
    apiKey: fetch_your_key

A standalone editor: Zed

In Zed you add an OpenAI-compatible provider through the agent settings (the Add Provider button) or in the settings file:

{
  "language_models": {
    "openai_compatible": {
      "zukko": {
        "api_url": "https://api.zukko.pro/v1",
        "available_models": [
          { "name": "gpt-6.1-sol", "display_name": "GPT-6.1 Sol", "max_tokens": 1000000 }
        ]
      }
    }
  }
}

Zed reads the key from its keychain or from an environment variable named after the provider ID: for zukko that's ZUKKO_API_KEY. By default Zed assumes such models don't accept image input.

How to choose

1. Where you work. Terminal — Claude Code, Codex, OpenCode. VS Code — Cline, Roo Code, Kilo Code, Continue. JetBrains — Cline, Kilo Code, Continue. A lightweight standalone editor — Zed.
2. Which models you need. If your main model is Claude, pick a client that uses the Anthropic format (Claude Code, Roo Code with the Anthropic provider, OpenCode): only there does the cache-read discount apply to Claude. In OpenAI-compatible clients all Claude input is billed at the input price. For GPT models caching shows up in both formats.
3. What it costs. Prices as of October 7, 2026, USD per 1M tokens input / output: claude-sonnet-5 — 0.4 / 2, claude-opus-5-5 — 0.8 / 4, gpt-6.1-sol — 0.27 / 1.32. A sample monthly estimate is in the article how much Claude Code costs per month↗. For the current list and prices, call GET https://api.zukko.pro/v1/models or open the Models page in the dashboard.
4. Test on your own tasks. Take two or three real tasks from your project and run them in two tools. Agents gather context differently, so token usage for the same task will differ.

One key works with all eight tools, but it's handier to create a separate key for each, with a daily limit and an allowed-models list. Create a key in the dashboard↗; Cline, Roo Code and Kilo Code are covered step by step in the VS Code extensions guide↗.

FAQ

Which Cursor alternative is closest in how it works?

If you want an agent inside VS Code, the closest are the Cline, Roo Code, Kilo Code and Continue extensions. If you're fine with a terminal, look at Claude Code, Codex and OpenCode.

Why do some tools need /v1 in the address and others don't?

Claude Code and Roo Code's Anthropic provider append /v1/messages themselves, so they need https://api.zukko.pro. OpenAI-compatible clients and OpenCode expect the address with /v1. A wrong address usually results in a 404.

Can I use Claude models in Codex?

Yes, Codex works with GPT and Claude models through the Responses API. GLM models don't work in Codex.

What if the tool says "model not found"?

Check the model ID: it must match what GET https://api.zukko.pro/v1/models returns. An unknown name gets 404 model_not_found. In Claude Code a common cause is background requests to claude-haiku-* when ANTHROPIC_DEFAULT_HAIKU_MODEL isn't set.