About
AI that doesn't say no.
77th Junction gives developers, researchers and writers one API key to uncensored language models, wired straight into the tools they already use.
Mainstream assistants refuse a surprising amount of ordinary work: security research on your own systems, reverse engineering, fiction with dark themes, frank medical or legal questions. Every refusal breaks your flow, and inside an agent it can derail a whole task.
We host models whose refusal behaviour has been removed and serve them behind a standard OpenAI-style API. Point Cursor at us and Agent mode keeps working instead of stopping to lecture.
Uncensored, not unsafe
Abliterated models answer what you ask, without refusals or moralising. A short list of hard limits keeps the service lawful.
Built for builders
OpenAI-compatible from day one, so Cursor, Cline, Continue and your own code work by changing two settings.
Fair, visible pricing
Prepaid credit from $1, charged per token. Every request appears in your log with its exact cost.
Private by default
We record token counts and cost for billing. We do not store your prompts or the replies.