How-to guides

Practical guides, each answering one concrete "how do I X". They assume the ground the tutorial covers (a project, a device, a model, a run) and go deeper on one problem at a time, in any order. Entries marked as coming are planned and land here as they are written.

Benchmarks and results

• Read benchmark results: the metrics a run returns, the per-sample evidence behind a score, and how to pull both through the API and the CLI.
• Compare two runs (coming): pick a baseline, read a regression, and share the comparison.

Definitions

• Write a job definition: the config.yaml shape, every key explained, a complete example, and how to apply it to a deployment.
• Write a service definition: the same treatment for a process the platform keeps running on a device.

Automation

• Set up MCP for Claude: give Claude Code or Claude Desktop the platform as a set of tools, over the deployment's own endpoint or the CLI's stdio server.
• Run benchmarks from CI (coming): an API key, clika-rt apply, and clika-rt benchmarks watch as a gate.

Administration

• Manage members and API keys: who is in the organization, who is in the project, and how an integration gets its own scoped credential.
• Activate a deployment license: activate a platform license, renew it, and issue the runtime credentials a ClikaRT project needs.

For the vocabulary these guides use, see Concepts. For the commands, see the CLI reference.