> ## Documentation Index
> Fetch the complete documentation index at: https://docs.minns.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Action suggestions

> The Policy Guide endpoint — ask 'What should the agent do next?'

This is the **Policy Guide** endpoint. Given the current context, it recommends the next best action based on past experiences and learned strategies.

## Query parameters

<ParamField query="context_hash" type="integer" required>
  Context fingerprint (computed from the current context).
</ParamField>

<ParamField query="last_action_node" type="integer">
  Last action node ID. Set to `null` to get suggestions from the start.
</ParamField>

<ParamField query="limit" type="integer" default="10">
  Maximum number of suggestions to return.
</ParamField>

## Response

Returns an array of `ActionSuggestionResponse` objects.

<ResponseField name="action_name" type="string">
  Suggested action name.
</ResponseField>

<ResponseField name="success_probability" type="number">
  Probability of success (`0.0`–`1.0`).
</ResponseField>

<ResponseField name="evidence_count" type="integer">
  Number of supporting past examples.
</ResponseField>

<ResponseField name="reasoning" type="string">
  Explanation for why this action is recommended.
</ResponseField>

```json theme={null}
[
  {
    "action_name": "check_seat_availability",
    "success_probability": 0.94,
    "evidence_count": 12,
    "reasoning": "After confirming the movie title, checking availability has a 94% success rate based on 12 past interactions."
  }
]
```

### Usage with the SDK

```typescript theme={null}
const suggestions = await client.getActionSuggestions(contextHash, lastActionNode, 3);

if (suggestions.length > 0) {
  const best = suggestions[0];
  console.log(`Recommended: ${best.action_name} (${(best.success_probability * 100).toFixed(0)}% success)`);
  console.log(`Reasoning: ${best.reasoning}`);
}
```