Updated November 15, 2026

Uncensored AI coding agent, local or in the browser

This uncensored AI coding agent works on your code with no extra filter from LU. Point it at a repository and describe the change: it plans, edits, runs the tests and hands you a pull request. Run it in the browser on LU Cloud from a 15 EUR credit pack, or free on your own PC.

For the agent on LU Cloud

Pick a plan for coding

Hosted
€19/month
€190/year · billed annually
900,000 credits/monthOne shared cloud wallet
  • About 31 typical agent runs on GLM 5.3
  • 11 Flash chat models without credits
Choose Hosted →
Hosted Max
€99/month
€990/year · billed annually
5,000,000 credits/month5.6× the Hosted wallet
  • About 175 typical agent runs on GLM 5.3
  • 12 Flash chat models without credits
  • GLM 5.3 Flash included
Choose Max →
450,000 credits
€15

About 15 agent runs on GLM 5.3, no subscription. The credits do not expire.

Buy credits →

Every plan and every pack opens the same catalogue; they differ in volume. A plan renews each month until you cancel it, a pack is paid once. Checkout opens on lu-labs.ai, and the payment page uses neutral product language. Full tables on the pricing page.

What the agent does

There are two places to run it, and one more way to use the same idea in chat.

Which models it runs on

LU Cloud runs the agent on 40 open chat models, and all of them can call tools. You pick the model per task. A cheap model is fine for a small fix. A large model is worth its price when the change touches many files or needs careful reasoning.

On your PC the agent uses the model you install. How large that model can be depends on your graphics card.

What one run costs

One typical run is a task that reads a few files, changes two or three and runs the tests. That is around fifteen steps. The table shows the credits for such a run on five models, and how many runs fit into one month of each plan.

ModelCredits per run, aboutRuns on HostedRuns on Hosted ProRuns on Hosted Max
DeepSeek V4.1 Flash 10175,900152398847
GLM 5.328,5003182175
DeepSeek V4 Pro 081336,4002464137
Qwen 3.8 Max48,7001848102
Kimi K3.1 Pro65,000133676

The runs per plan are alternatives, not a sum. One wallet pays for chat, code, images and video, so every run on one model leaves less for the others. These are estimates, not a quota. A run that goes in circles costs credits like one that succeeds. A large repository or a long session uses more, and a cached prompt makes a run cheaper. For every model, see the full cost table on LU Labs.

Plans and packs for coding

Plans also include Flash models that cost no credits in app sessions, on an active paid plan: 11 on Hosted, 12 on Pro and Max, up to 500,000 combined input and output tokens per account per UTC day. GLM 5.3 Flash uses credits on Hosted and is included on Pro and Max. API keys and accounts without an active subscription always use credits.

Checkout opens on lu-labs.ai. Compare the plans on LU Labs.

Where uncensored fits for coding

Most code is not sensitive. Any capable model will write a form validator or a database migration. The word uncensored matters at the edges, where cautious assistants refuse work that is legitimate. Typical cases:

On LU Cloud, hosted text has no extra filter from LU. The models themselves differ. Some were trained to decline more than others, and no model writes everything you ask. If one refuses a legitimate task, try another from the picker. The No Filter mark comes from a writing test, not a coding test, so treat it as a hint: 13 of the 36 cloud chat models we measured answer in full without refusing, and only those carry the No Filter mark.

Illegal use stays illegal. Attacking systems you do not own or spreading malware is not what this is for, and the Terms of Service apply.

Where your code goes

On LU Cloud the code is processed on GPU partners. LU Labs never trains on it and never sells it. In the browser the agent only sees the repositories you chose.

If your code must not leave your machine, run the desktop app with a local model. That route is free.

The local route

Locally Uncensored is free and open source under AGPL-3.0, for Windows, macOS on Apple Silicon and Linux. It runs the agent on folders on your own machine, with the model you install, and needs no account and no credits. The limit is your hardware: the largest models will not fit on a home PC.

Would you rather use the hosted models from your own editor, script or agent? The uncensored AI API page shows how to call them from your own tools.

How to start

  1. Open the LU Labs pricing page and pick a credit pack or a plan.
  2. Open the Code tab, connect GitHub and choose which repositories the agent can see.
  3. Pick a model, describe the change and review the pull request it opens.

Questions people ask

How many credits does one agent run cost?

It depends on the model. A typical run reads a few files, changes two or three and runs the tests, around fifteen steps. That costs about 5,900 credits on DeepSeek V4.1 Flash 1017, about 28,500 on GLM 5.3 and about 65,000 on Kimi K3.1 Pro. A large repository or a long session uses more, and a run that goes in circles costs credits like one that succeeds.

Is the uncensored AI coding agent free?

The local route is free: the desktop app runs the agent with a local model on your own PC. The hosted route on LU Cloud has no free tier. It starts with a credit pack from 15 EUR for 450,000 credits, with no subscription, and the credits do not expire.

Which plan should I take for coding?

Hosted at 19 EUR a month, with 900,000 credits, to try the agent and for occasional tasks. Hosted Pro at 49 EUR, with 2,350,000 credits, for a few agent tasks every working day. Hosted Max at 99 EUR, with 5,000,000 credits, for daily heavy use on the largest models. One wallet pays for chat, code, images and video.

Which models can the agent use?

All 40 open chat models on LU Cloud, because all of them can call tools. In our cost table DeepSeek V4.1 Flash 1017 is the cheapest per run and Kimi K3.1 Pro the most expensive. On your own PC the agent uses the model you install.

Does my code leave my machine?

On LU Cloud, yes: the code is processed on GPU partners. LU Labs never trains on it and never sells it. In the browser the agent only sees the GitHub repositories you choose, and changes arrive as pull requests. If the code must stay on your computer, use the desktop app with a local model.

Is it really uncensored?

LU adds no extra filter to hosted text. The models still differ, and no model writes everything you ask. It helps with legitimate work that cautious assistants refuse, such as security research on your own systems, reverse engineering for interoperability or malware analysis in a lab. Illegal use stays illegal.

Read next

40 open chat models with tool calls, runs from about 5,900 credits, credits from 15 EUR with no subscription.

See Hosted Pro Buy a credit pack

Prefer everything offline? The desktop app is free, AGPL-3.0, Windows, Linux and Mac.

Current Hosted offer

Hosted includes 11 Flash models; Pro and Max include 12. In app sessions they cost no credits, up to a daily limit per account. GLM 5.3 Flash uses credits on Hosted and joins the daily allowance on Pro and Max. API keys always use credits.

EUR 19/month, 900,000 shared monthly credits: up to 900 Neta Lumina Uncensored images or 90 five-second LTX 2.3 Uncensored clips. Every output spends the same wallet.

Each maximum spends the entire shared monthly wallet on one model. These are alternatives, not separate allowances. Input-image generation, character training, LoRA routes and additional operations cost extra. Uncensored video uses an input image and five-second clips. Failed jobs and changing catalog prices can affect results.

Image models and exact quantities · Video models and exact quantities · Venice comparison · Current facts JSON