Use local AI agents in Cursor and Continue.dev

3 min read Original article ↗

For developers using Cursor or Continue.dev

Locaible exposes an OpenAI-compatible API on 127.0.0.1:11435. Point Cursor or Continue.dev at it and your AI runs entirely on your machine, with pre-tuned coding agents from the Marketplace.

Your code never leaves your machineZero token billGDPR-safe for client repos

5-minute walkthrough — see Locaible behind Cursor end-to-end.

Pre-tuned coding agents

Atlas (legal), Code Reviewer (OWASP-aware), Tech Support (L2/L3 runbooks) appear as 'models' in Cursor's dropdown. Each carries its own system prompt + RAG.

Per-project agents

One client repo with strict data-handling rules? Spin up an agent that knows your conventions and feed it your style guide + arch docs. Switch via the Cursor model menu.

Compliant by construction

Cursor Pro sends snippets to Anthropic/OpenAI. With Locaible behind it, the chat traffic stops at 127.0.0.1. Defensible to your CISO and any DPO audit.

Setup — 3 minutes

Same flow for Cursor and Continue.dev. Substitute your endpoint and you're done.

  1. 1

    Install Locaible and create a coding agent

    Pick a model that fits your hardware — qwen3:8b for fast iteration, qwen3:14b for serious refactors, deepseek-coder:33b if you have a 48 GB rig. Locaible installs Ollama for you on first launch.

    Get the desktop app

  2. 2

    Start the local OpenAI API in Locaible

    Settings → Local API → Start. Locaible binds a server on 127.0.0.1:11435 that speaks the OpenAI spec and exposes your agents as virtual models.

    # Verify it's up

    $ curl http://localhost:11435/v1/models

    { "object": "list", "data": [

    { "id": "atlas", "object": "model", "owned_by": "locaible" },

    { "id": "code-reviewer", "object": "model", ... }

    ] }

  3. 3

    Point Cursor (or Continue) at the endpoint

    1. Cmd/Ctrl + , → Models
    2. Toggle OpenAI API Key on. Paste any non-empty string as key.
    3. Click Override OpenAI Base URL, paste:
    4. In the model dropdown of the chat: type atlas (or your agent name) and select it.

    ⊕

    Continue.dev

    (full-local)

    1. Install Continue VS Code extension.
    2. Edit ~/.continue/config.json — add:
    {
  "models": [
    {
      "title": "Locaible · Atlas",
      "provider": "openai",
      "model": "atlas",
      "apiKey": "noop",
      "apiBase": "http://localhost:11435/v1"
    }
  ]
}

    Continue also supports local Tab autocomplete via Ollama — Cursor's autocomplete stays on their cloud.

What works, what doesn't

Brutal honesty so you don't discover the gaps after switching.

Bottom line: Cursor + Locaible is the right call if you want to keep Cursor's UX and only offload the heavy chat/edit traffic to your machine. If full-local is a hard constraint (regulated industry, defence, finance), Continue.dev + Locaible is the path.

Pro tip: lock Cursor to Locaible per project

Drop a .cursorrules file at your repo root telling Cursor (and your future self) which Locaible agent to use. Especially useful for client work — you can't accidentally send their code to OpenAI.

# .cursorrules — committed to the repo
# All AI assistance for this codebase MUST go through the local
# Locaible endpoint. Selected model: atlas-acmecorp.
#
# Endpoint: http://localhost:11435/v1/chat/completions
# Reason: client NDA + GDPR Art. 28 forbid third-party processors
# for source code. Locaible runs 100% on the dev's machine.

Get your Cursor running on your own machine

Free tier covers 3 agents and 1 device. Enough to try every step on this page.

Download LocaibleGet Continue.dev