Works withyour agent
Same models, same balance, whichever harness you run. Pick yours for the exact config.
A page for each harness
The panel above writes the config for whichever model you pick. Each harness also has a page of its own, with the steps in order and the things that usually go wrong.
opencodeA provider block in opencode.json, with the real context limits so sessions are not compacted early.Hermes AgentPoint Hermes Agent at an uncensored, abliterated model line with a custom endpoint: the config.yaml and .env to use, and which model suits a long-running agent.Claude CodeRun Claude Code against uncensored and abliterated models by setting three environment variables. The endpoint speaks the Anthropic format Claude Code sends.CodexAdd an uncensored model provider to Codex CLI in config.toml: the exact block, the environment variable for the key, and the model ids to use.SillyTavernConnect SillyTavern to uncensored and abliterated models as a custom Chat Completion source. The exact fields to fill in, and why no jailbreak prompt is needed.
The base URL and the key
Every harness needs the same two things. The base URL is https://uncensoredopencode.com/v1 for anything that speaks the OpenAI format, and https://uncensoredopencode.com for Claude Code, which speaks the Anthropic format and adds the path itself. The key starts with sk-nc- and comes from the dashboard after you sign in and add credit.