Skip to content
Outrun
Open dashboard
UNCENSORED · VS CODE EXTENSION

Uncensored ModelsRoo Code (VS Code).

Privacy first · ZDR explained

Outrun keeps prompts and responses out of operational request logs across every model.

ZDR means zero data retention. A model’s ZDR badge identifies an additional processor routing control; processor limits and audit exceptions still apply.

Your connected app may save history. Outrun Playground text chats are temporary by default; account saving is optional. Media and Workspace use separate storage rules.

Privacy and retention details

Connect Roo Code with your Outrun key, choose an uncensored model, and send a short test. The same connection values work in other OpenAI-compatible apps.

Before you start

Have your API key and some Outrun credit ready. Keep your key private.

Get your API keyCheck balanceChoose a model
Download VS CodeGitHub Repository
Connection settingsShow
Connection settings
API Provider: OpenAI Compatible
Base URL: https://api.outrun.to/v1
First Model ID: glm-5.3-abliterated
GLM Context Window: 1000000
GLM Max Output Tokens: 65536
Fast Abliterated Model ID: deepseek-v4.1-flash-abliterated-fast
DeepSeek Abliterated Model ID: deepseek-v4.1-flash-abliterated

Get your key from Outrun → API keys and enter it privately in the app or local settings. Test with “Reply with OK only.” Tests use credit; check Activity before retrying.

Setup Steps

  1. 1
    Install VS Code

    Download Visual Studio Code (VS Code) from https://code.visualstudio.com/download. VS Code is the desktop app that Roo Code runs inside. Choose Windows, macOS or Linux, open the downloaded installer, follow the installation prompts, then open VS Code.

    Download VS Code
  2. 2
    Install Roo Code

    Inside VS Code, click Extensions (the four-squares icon on the left), search for Roo Code, and click Install on the Roo Code extension. Roo Code is an AI assistant extension inside VS Code; it is not a separate desktop app. After installation, open its kangaroo icon on the left.

    Open Roo Code extension page
  3. 3
    Open provider settings

    Open Roo Code’s kangaroo icon on the left. Click its settings gear. Choose OpenAI Compatible as API Provider.

  4. 4
    Enter the URL and key

    Base URL: https://api.outrun.to/v1. API Key: paste your own key into that field, never into chat. Leave the URL exactly as shown; do not add /chat/completions.

  5. 5
    Choose your first model

    Use glm-5.3-abliterated for a first test. To try the others later, save a separate profile with glm-5.3-flash-abliterated or deepseek-v4-pro-uncensored as its Model ID. Additional available models: deepseek-v4.1-flash-abliterated-fast (DeepSeek V4.1 Flash Abliterated Fast), or deepseek-v4.1-flash-abliterated (DeepSeek V4.1 Flash Abliterated). Copy the exact Model ID into a separate profile; the GLM capability values below apply only to GLM.

  6. 6
    Set model capacity

    For GLM Abliterated and Flash Abliterated, set Context Window to 1000000 and model Max Output Tokens to 65536. These are different limits. Never set output tokens to 1000000. Turn Image Support and Computer Use off. Enable streaming and Native tool calling.

  7. 7
    Set a small test limit

    If Roo shows a separate Include Max Tokens/request-limit setting, enable it and set 1024 for a first test. This controls the test response length; it is separate from model capacity.

  8. 8
    Save and test

    Click Save and select this profile for a new task. Start a new chat and send: Reply with OK only. A short reply with no connection error means the first test worked. This request uses Outrun credit; check Activity if it is interrupted before trying again.

Manual settings and examplesShow
Optional · key variable for the cURL test
# Only needed for the cURL test below (Roo Code stores its key in its own settings).
export OUTRUN_API_KEY="KEY"

Replace KEY with your own key only in your local app, editor or terminal. Never send that key to an AI chat. Copy one command block at a time.

Configuration Template

Config Settings
Roo Code settings
API Provider: OpenAI Compatible
Base URL: https://api.outrun.to/v1
API Key: enter your key locally
Model ID: glm-5.3-abliterated
# GLM Abliterated / Flash Abliterated model capabilities
Context Window: 1000000
Max Output Tokens: 65536
# Separate request limit, if shown
Include Max Tokens: on
Request Max Tokens: 1024 for a first test
Streaming: on
Tool Protocol: Native
Image Support: off
Computer Use: off
# Other new models: glm-5.3-flash-abliterated, deepseek-v4-pro-uncensored
# Never use context size as the output-token limit. Save and select this profile.
# Additional Model IDs (separate profiles): deepseek-v4.1-flash-abliterated-fast, deepseek-v4.1-flash-abliterated
# The GLM capability values above apply only to GLM; select limits appropriate to the chosen model.
Optional: test with a script (uses credit)Show
Test cURL
curl https://api.outrun.to/v1/chat/completions \
  -H "Authorization: Bearer $OUTRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5.3-uncensored","messages":[{"role":"user","content":"Reply with OK only."}],"max_tokens":1024}'

If something goes wrong

Key missing or rejected (401)

Copy your current key from API keys. Paste it only into the app’s key field or your local key variable, then save and reopen the app.

Not enough available credit (402)

Check Balance and the key’s spending cap. Another request may have temporarily reserved credit; read the message before adding funds.

Wrong URL, model or limit (400/404)

Copy the exact values above. Do not add an operation such as /chat/completions to a Base URL. For GLM, context is 1000000 and maximum output is 65536.

Busy or rate limited (429)

Wait for the active request to finish or the rate limit to reset. Avoid repeatedly submitting the same request.

Interrupted request or uncertain charge

Check Activity first. Keep the request ID for support; never include your key in a message or screenshot.

Model choices and service notes

Available model IDs: glm-5.3-uncensored, glm-5.3-flash-uncensored, glm-5.3-abliterated, glm-5.3-flash-abliterated and deepseek-v4-pro-uncensored. Abliterated variants are designed to reduce refusals; responses can still vary. GLM Abliterated and Flash Abliterated were verified with Roo Code native tool calls. If a request returns 402, check your account balance and key spending cap. For 429, read the error: wait for the active request to finish or the rate limit to reset. If a paid request is interrupted or its outcome is uncertain, check Activity before trying again; do not automatically repeat it. Choose an available model from Models & Pricing; paused models cannot be used. MiMo model IDs: mimo-v2.6-flash-uncensored-thinking, mimo-v2.6-flash-abliterated, mimo-v2.6-flash-uncensored. MiMo reasoning tokens are output tokens and use the output rate. Model/tool capabilities depend on the chosen model; do not use the context window as an output-token limit.

Contact @outrunsupport