The frontier of
refusal-less models,
with absolutely zero data retention
We never store your prompts or the answers, and nobody here reads them. One unified API endpoint. MagmaRouter (Magma Router) is an uncensored LLM API for coding agents, security research and open-ended conversations. Access uncensored AI models and abliterated LLMs through one gateway. Connect Xploded models to Claude Code, Cursor, Codex, Cline or opencode through OpenAI-compatible and Anthropic-compatible endpoints.
Built for the work other APIs turn away
Uncensored Frontier Models
Frontier-grade open models with the refusal taken out of the weights, for reasoning, coding, research and agentic work. No moderation layer sits in front of them.
What you get
Compatible Endpoints
OpenAI and Anthropic compatible endpoints that fit the workflow you already have. Change only the base URL and the API key.
Works with
Claude Code
Cursor
opencode
ClineHard Spend Caps
Every key carries a lifetime or monthly cap it can never go over. The worst case is reserved before a request leaves, so a runaway agent stops at the limit.
Built for Security Work
Red teaming, penetration testing, malware analysis, exploit research, trust and safety, and the coding agents that do it all.
Seven models, one base URL
Point any OpenAI-compatible tool at https://api.magmarouter.com/v1, or anything built for Anthropic at https://api.magmarouter.com, and use one of these model ids. Every one calls tools and reasons.
| Model | Model id | Context | Images |
|---|---|---|---|
 GLM 5.3 Xploded | glm-5.3-xploded | 1.05M tokens | No |
| GLM 5.3 Flash Xploded | glm-5.3-flash-xploded | 1.05M tokens | Yes |
 DeepSeek V4.1 Flash 1017 Xploded | deepseek-v4.1-flash-1017-xploded | 1.05M tokens | Yes |
| DeepSeek Flash Lite Xploded | deepseek-v4.1-flash-lite-1017-xploded | 512k tokens | Yes |
 MiMo V2.6 Flash Xploded | mimo-v2.6-flash-xploded | 1.05M tokens | Yes |
 Qwen3.8 27B Xploded | qwen3.8-27b-xploded | 1.05M tokens | Yes |
| Qwen3.8 27B Ultrafast Xploded | qwen3.8-27b-fast-xploded | 524k tokens | Yes |
Questions people ask
What is MagmaRouter?
MagmaRouter is an uncensored and abliterated LLM API. It serves state of the art models that do not refuse, behind one OpenAI and Anthropic compatible endpoint, with spend caps on every key.
Do you store or read my prompts?
No. MagmaRouter keeps zero data retention. What you send goes to the model and the answer comes straight back, and neither one is written to our database, our logs or disk. The request log keeps only the model, token counts, cost and timing, so there is nothing for us or anyone else to read.
Which API gives uncensored LLM access?
MagmaRouter is built for exactly that. One API key reaches uncensored and abliterated builds of GLM 5.3, DeepSeek V4.1 Flash 1017, MiMo V2.6 Flash and Qwen3.8 27B, and every model on the line has tool calling, so it can drive a coding agent.
What does abliterated mean?
An abliterated model has had its refusal behaviour removed from the weights themselves, rather than being talked around with a prompt. It answers instead of declining.
Does it work with Claude Code, Cursor and the OpenAI SDK?
Yes. Point any OpenAI compatible tool at https://api.magmarouter.com/v1, or anything built for Anthropic at https://api.magmarouter.com. Claude Code, Cursor, Cline and opencode all work, and the docs walk through each one.
How do I pay for it?
Add credit by card or crypto and it is drawn down as you use it. There is no subscription, and credit never expires.
Point your harness at it. Start in a minute.
Try it in the chat first, or follow the docs to point Claude Code, Cursor or opencode at it.