CLI commands
Every command in L80 1.0. L80 <command> --help lists every flag; the linked guide covers each in depth.
Benchmark and publish
| command | does | needs |
|---|---|---|
| L80 betterbench --endpoint <url> --model <m> | Run BetterBench against an OpenAI-compatible server and publish the run. | Needs a token |
| L80 betterbench --results <results.json> | Publish an existing BetterBench results file. | Needs a token |
| L80 betterbench … --dry-run | Run or read and prepare the page without publishing. | Free, no token |
| L80 publish <file.json> [--template <id>] | Publish a hand-written payload. Prints the URL. | Needs a token |
| L80 templates list | The template ids the server accepts right now. | Free, no token |
| L80 history | The pages this key has published, newest first, with their trust tier. | Needs a token |
Verification
| command | does | needs |
|---|---|---|
| L80 verify check <results.json | url> | The checks a review applies, run locally. Exit 0 if the run qualifies, 4 if not. | Free, no token |
| L80 verify request <url> [--note …] | Ask Launch80 to verify a page this key published. | Needs a token |
| L80 verify status <url> | A page's latest request and its decision. | Needs a token |
| L80 verify mine | Your requests and their decisions. | Needs a token |
| L80 verify queue [--status …] | Other keys' requests, oldest first. | Needs a token |
| L80 verify show <id> | What to reproduce: model, engine, hardware, settings, published figures. | Needs a token |
| L80 verify reproduce <id> --endpoint <url> | Rerun BetterBench with the published settings on your hardware and compare. | Needs a token |
| L80 verify approve <id> [--note …] | Stamp a page verified. Never your own request. | Verifier keys |
| L80 verify reject <id> --reason … | Decline a request. The publisher sees the reason. | Verifier keys |
Tools
| command | does | needs |
|---|---|---|
| L80 vast <command> [args] | Rent a Vast.ai GPU, serve a model, benchmark it, publish, tear down. | Free, no token |
| L80 pwdet scan <model> [--json] [--publish] | Check weights for tampering. Publishing the report needs a token. | Free, no token |
| L80 dashboard [tui|serve] | GPU and inference telemetry. | Free, no token |
| L80 tools list|check|prefetch|path|run | Show, check, fetch and run the pinned Python tools. | Free, no token |
| L80 tui | The full-screen console for all of the above. | Free, no token |
Setup
| command | does | needs |
|---|---|---|
| L80 skills print [name] | Print a bundled agent skill. | Free, no token |
| L80 skills link --target claude-code [--name …] | Install a skill into Claude Code. | Free, no token |
| L80 auth status | The resolved server and a masked fragment of the token. | Free, no token |
| L80 doctor | Check connectivity, config, the token and every tool. | Free, no token |
| L80 update [--check] [--json] | Update L80, its pinned tools and installed skills. | Free, no token |
| L80 version | Print the version. | Free, no token |
Global flags
| flag | does |
|---|---|
| --json | Machine-readable output, on most commands. |
| --api-base <url> | Server override for any network command. Same as L80_API_BASE. |
| --help | Every flag for a command. |
Flags may come before or after a file argument. Errors print a code, a message and a remedy: line; see Errors and exit codes.