ALL
HARNESSES
WITH NAN.
Run any supported coding agent with NaN through nan-harness. One command checks compatibility, discovers your available models, prepares the connection, and supervises the process.
$ curl --proto '=https' --tlsv1.2 -m 30 -fsSL https://github.com/DavidLMS/nan-harness/releases/latest/download/install.sh | shPS> irm https://github.com/DavidLMS/nan-harness/releases/latest/download/install.ps1 | iex$ curl --proto '=https' --tlsv1.2 -m 30 -fsSL https://github.com/DavidLMS/nan-harness/releases/latest/download/install.sh | sh
The model moves.
Your workflow stays.
NaN provides the models. Your agent provides the workflow. nan-harness connects them across every supported harness while you keep the client and workflow you already use.
SEE THE WORKFLOW$ nanh opencode
→ recommended: checks, routes and supervises OpenCode
$ nanh config opencode
→ advanced: writes persistent OpenCode configuration
$ opencode
→ direct launch: uses copied credentials and a model catalog
_
Run it through nan-harness.
One command.
Use nanh <harness> for everyday launches. It checks compatibility, discovers your available NaN models, prepares any required bridge, and supervises the process. Use nanh config <harness> only when you need to start the harness directly; refresh that configuration when your saved key or model catalog changes.
Telemetry off.
Consent on.
Telemetry is off by default. If enabled, nan-harness sends sanitized diagnostics and minimal usage data. It never includes prompts, model output, arguments, paths, or credentials. A model-specific diagnostic may include the NaN model ID.
nanh telemetry onWORTH
KNOWING.
What is nan-harness?+
The recommended way to use supported coding agents with models from your NaN community subscription. The project command is nan-harness; nanh is its short alias.
Does it replace my agent?+
No. You keep the original agent, profile and workflows. nan-harness manages its launch and NaN connection without replacing the client or writing persistent provider settings.
When should I use nanh config?+
Use it only when you need to start a supported harness with its normal command. For everyday use, prefer nanh <harness>; native setup must be refreshed when your saved key or model catalog changes and is unavailable for Claude Code, Codex, and fx.
How are models discovered?+
nanh <harness> reads the current NaN catalog at launch. Advanced native setup writes a catalog snapshot into the harness, so run nanh config <harness> --refresh when your available models change.
Can I request another harness?+
Yes. If a harness you use is missing, open an issue and tell us which one you would like to see supported.