Run rakuvo serve -h for the authoritative list. The same settings can be pushed from the control plane (see API reference) and are validated there.
#Gateway
| Flag | Meaning |
|---|
-listen | Address to listen on. A non-loopback address requires -token-env. |
-token-env | Name of the environment variable holding the bearer token clients must send. |
-trace | File to write one JSON line per model call. |
-upstream | Base URL of the model provider (default OpenAI). |
-upstream-key-env | Environment variable holding the provider key. |
-max-tokens-field, -omit-temperature | Provider compatibility switches. |
| Flag | Meaning |
|---|
-tool-policy | Always add the tool-use instruction. |
-tool-policy-auto | Add it only for models where it was measured to help. |
-guard, -guard-pass-through | Tool-call guard on or off; let a failing call through instead of refusing it. |
-verify-model | Judge model. Empty turns Verify off. |
-verify-mode | veto, gate or both. |
-verify-scope | single (default) or all. |
-judge-prompt-file | Replace the judge's system prompt. |
#Context
| Flag | Meaning |
|---|
-context-keep-observations | Keep this many of the latest tool results in full; mask older ones. |
-context-mask-step | Mask in jumps of this many results so the cached prompt prefix stays stable. |
-context-mask-within-turn | Also mask inside the latest user turn (long single-request agent runs). |
-context-max-tokens | Token budget for the whole request; drops the oldest whole turns when over. |
-context-summarize-model, -context-facts-model | Models for summaries and the key-facts note. |
-context-keep-relevant, -context-prune-tools | Keep results relevant to the latest message; prune the tool list. |
#Loop, model policy, effort, memory
| Flag | Meaning |
|---|
-loop-guard, -loop-warn-only | Turn the loop guard on; only nudge, never stop. |
-loop-max-tool-calls, -loop-allow-repeat | Hard cap on tool calls; tool names allowed to repeat. |
-model-cheap, -model-strong | The two tiers of the cascade. |
-model-escalate-on-no-call | Also escalate when the cheap model answered in text. |
-reasoning-effort, -reasoning-effort-escalated, -reasoning-force | Reasoning effort for reasoning models, and a higher level for retries. |
-memory-dir | Directory for per-user memory. Required to enable Memory. |
-memory-model, -memory-embed-model, -memory-select-model | Models for extraction, embeddings and evidence selection. |
-memory-top-k, -memory-scope | How many facts to recall; how users are separated. |
#Control plane
| Flag | Meaning |
|---|
-config-from-api | Pull layer settings from the control plane instead of flags. |
-api-url | Control-plane URL. |
-api-key-env | Environment variable with your Rakuvo API key (default RAKUVO_API_KEY). |
-config-poll | How often to check for new settings. |
Layer flags conflict with -config-from-api: settings come from one place.