Locaible
local AI agents for Cursor with pre-tuned marketplace/commu
Last verified:
What is Locaible?
Locaible is a desktop application that runs pre-tuned AI coding agents directly on your machine and exposes them to Cursor (and Continue.dev) through a local, OpenAI-compatible API endpoint at 127.0.0.1:11435. Instead of sending your codebase to hosted cloud models like Anthropic or OpenAI, you point Cursor at this localhost endpoint, ensuring your code never leaves your machine. This provides a privacy-first, GDPR-safe alternative for developers working under compliance constraints or on client repos with strict data-handling rules.
The tool ships with pre-tuned specialist agents including Atlas (a legal-focused agent), Code Reviewer (OWASP-aware security reviewer), and Tech Support (backed by L2/L3 runbooks). Each agent carries its own system prompt and RAG (retrieval-augmented generation). Beyond these built-in options, Locaible offers a Marketplace for additional agents and lets you build custom per-project agents loaded with your own style guides and architecture documentation, which you can switch between via Cursor's model dropdown menu.
Under the hood, Locaible runs local models through an Ollama-style stack, supporting models from qwen3:8b for fast iteration up to qwen3:14b for serious refactors, and deepseek-coder:33b for users with 48 GB GPU rigs. Locaible installs Ollama automatically on first launch. Because it speaks the standard OpenAI API format, the same local endpoint works with any tool accepting a custom base URL, not just Cursor.
Locaible is designed for developers and creators building with AI who work in regulated industries (defense, finance, healthcare), on client projects with NDAs, or anyone prioritizing data privacy. It's particularly valuable for teams needing defensible AI usage against CISO audits or DPO compliance checks, as all chat traffic stops at localhost rather than reaching cloud providers.
Locaible pricing
Pricing model: Freemium
Free tier covers 3 agents and 1 device, which is enough to try all setup steps. A paid tier exists for more agents and devices (specific pricing details not disclosed on the page). No token bills since everything runs locally on your machine.
Locaible pros
- Code never leaves your machine - complete privacy
- Zero token bills - no cloud API costs
- GDPR-safe for client repos and regulated work
- Compliant by construction - defensible to CISO/DPO
- Pre-tuned agents shipped ready-to-use (Atlas, Code Reviewer, Tech Support)
- Each agent has custom system prompt + RAG built-in
- Marketplace available for additional pre-tuned agents
- Custom per-project agents with your style guides
- Switch agents via Cursor's model dropdown menu
- Supports multiple models (qwen3:8b, qwen3:14b, deepseek-coder:33b)
- Locaible installs Ollama automatically on first launch
- OpenAI-compatible API works with any custom base URL tool
- Free tier covers 3 agents and 1 device
- Works with both Cursor and Continue.dev
- Local chat with full code context
- Local inline edit (Cmd+K style) functionality
- Per-project .cursorrules file prevents accidental cloud sends
- No third-party processors for source code
- Hardware-adaptive model selection based on GPU
- Local Tab autocomplete in Continue.dev
Locaible cons
- Tab autocomplete in Cursor still uses cloud (not local)
- Full repo codebase indexing in Cursor is cloud-based
- Agent/autonomous edits are limited in both Cursor and Continue.dev
- Cursor + Locaible only provides partial local code protection
- Requires sufficient GPU hardware for larger models (48 GB for 33b)
- Model quality trade-off compared to cloud agents exists
- Limited autonomous editing capabilities
- Not full-local for Cursor (only Continue.dev is truly full-local)
- Setup requires manual endpoint configuration in Cursor
- Ollama dependency adds installation overhead
Frequently asked questions about Locaible
What is Locaible and what does it do?
Locaible is a desktop app that runs pre-tuned AI coding agents on your own machine and exposes them to Cursor through a local, OpenAI-compatible API endpoint at 127.0.0.1:11435. Instead of sending your codebase to hosted cloud models, you point Cursor at localhost, ensuring your code never leaves your machine while still getting agent workflows.
Which models does Locaible support?
Locaible runs local models through an Ollama-style stack, supporting options from qwen3:8b for fast iteration, qwen3:14b for serious refactors, up to deepseek-coder:33b if you have a 48 GB GPU rig. Locaible installs Ollama for you automatically on first launch.
What pre-tuned agents come with Locaible?
Locaible ships with pre-tuned specialists including Atlas (a legal-focused agent), Code Reviewer (OWASP-aware security code reviewer), and Tech Support (backed by L2/L3 runbooks for technical support). Each agent carries its own system prompt and RAG.
Is Locaible GDPR-safe for client work?
Yes, Locaible is GDPR-safe for client repos. Your code never leaves your machine, and there are no third-party processors for source code. This is defensible to your CISO and any DPO audit, especially important for client work with NDAs and GDPR Article 28 compliance.
How do I set up Locaible with Cursor?
First install Locaible and pick a model. Then go to Settings → Local API → Start in Locaible to bind the server. In Cursor, press Cmd/Ctrl + , → Models, toggle OpenAI API Key on (paste any non-empty string), click Override OpenAI Base URL and paste http://localhost:11435/v1, then select your agent (like 'atlas') from the model dropdown.
Does Locaible work with Continue.dev too?
Yes, the same flow works for Continue.dev. Install the Continue VS Code extension and edit ~/.continue/config.json to add a model configuration with title 'Locaible · Atlas', provider 'openai', model 'atlas', apiKey 'noop', and apiBase 'http://localhost:11435/v1'. Continue also supports local Tab autocomplete via Ollama.
What features work locally vs cloud with Cursor + Locaible?
With Cursor + Locaible: Chat with code context is local, inline edit (Cmd+K) is local, switching between agents is local. However, Tab autocomplete is cloud-based, full repo codebase indexing is cloud-based, and agent/autonomous edits are limited. Your code only partially stays on the laptop. Continue.dev + Locaible is fully local for all features.
When should I choose Cursor + Locaible vs Continue.dev + Locaible?
Cursor + Locaible is right if you want to keep Cursor's UX and only offload heavy chat/edit traffic to your machine. If full-local is a hard constraint (regulated industry, defense, finance), Continue.dev + Locaible is the path since it provides complete local protection for all features including Tab autocomplete and codebase indexing.
Can I create custom agents for specific projects?
Yes, you can spin up per-project agents that know your conventions. Feed them your style guide and architecture docs, then switch between them via the Cursor model menu. This is especially useful for client repos with strict data-handling rules.
How do I prevent accidentally sending client code to OpenAI?
Drop a .cursorrules file at your repo root telling Cursor which Locaible agent to use. This file should specify the endpoint (http://localhost:11435/v1/chat/completions), selected model, and reasoning (client NDA + GDPR Art. 28). This prevents accidentally sending client code to OpenAI and is committed to the repo.