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

# What are you trying to do?

> Use-case-first routing: map an intent to the exact tool sequence.

Pick the surface by **what the user is trying to do**, then run the tool sequence for that intent. When two surfaces could answer, prefer the one that holds the actual answer — numbers live in Data, durable truth lives in Knowledge.

## Intent → tool sequence

| Intent                                                                                                     | Tool sequence                                                                          |
| ---------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- |
| **Exact business data / numbers** — products, stock, orders, prices, sales, counts, procurement, customers | `data_view_list` → `data_query`                                                        |
| **Durable company truth** — SOP, strategy, policy, brand rules                                             | `knowledge_search` → `knowledge_get`                                                   |
| **Episodic / contextual recall** — past events, user-specific facts                                        | `memory_search` → `memory_get`                                                         |
| **What an agent did before**                                                                               | `agent_search_runs` → `agent_run_get`                                                  |
| **Reusable AI capability**                                                                                 | `skill_list` → `skill_get`                                                             |
| **Multi-step automation**                                                                                  | the `mission_*` family — see [Mission Recipes](/api/ai/mission-recipes)                |
| **Market / competitive intelligence**                                                                      | the `vertical__vertical_*` family — see [Vertical Playbook](/api/ai/vertical-playbook) |

<Warning>
  Never answer a numeric operational question (how many, how much, current stock, this month's sales) from Knowledge when Data exists. Knowledge holds policy and prose; Data holds the live records and numbers. Route counts and amounts to `data_query`.
</Warning>

<Note>
  For Knowledge, read the **full** content with `knowledge_get` before you decide — don't answer on the search-result intro alone. The intro is a snippet; the governing detail is usually in the body.
</Note>

For market intelligence, use the Vertical tools rather than generic web search — they return grounded, brand-specific signals with coverage you can check. See the [Vertical Playbook](/api/ai/vertical-playbook).

## Concrete questions

Natural asks mapped to the sequence that answers them:

| The user asks…                           | Do this                                                              |
| ---------------------------------------- | -------------------------------------------------------------------- |
| "Find a product"                         | `data_view_list` → `data_query`                                      |
| "Change a product field"                 | `data_row_update` (`dry_run: true` first)                            |
| "Find an SOP"                            | `knowledge_search` → `knowledge_get`                                 |
| "Update an SOP"                          | `knowledge_update` (update in place — don't duplicate)               |
| "Create an agent"                        | `department_list` + `skill_list` + `grantable_list` → `agent_create` |
| "Run an agent"                           | `agent_run` (human-gated — creates a pending approval)               |
| "Why did my mission fail?"               | `mission_status` → `mission_run_trace_get`                           |
| "Pause my mission"                       | `mission_pause`                                                      |
| "Compare two competitors"                | `vertical__vertical_compare`                                         |
| "What does the market say about Brand X" | `vertical__vertical_resolve` → `vertical__vertical_card`             |

<Note>
  The discovery step is not optional. "Find a product" starts with `data_view_list` because you need the real view slug and column names before `data_query` — the same holds for agents (departments, skills, grants), missions (IDs), and vertical (resolve the brand). See [Discover before guessing](/api/ai/operating-contract#discover-before-guessing).
</Note>

Deeper routing for the two orchestration surfaces:

<CardGroup cols={2}>
  <Card title="Mission Recipes" icon="diagram-project" href="/api/ai/mission-recipes">
    Explain, start, diagnose, control, edit, and compare missions.
  </Card>

  <Card title="Vertical Playbook" icon="chart-line" href="/api/ai/vertical-playbook">
    Ground market answers in retrieved, brand-specific signals.
  </Card>
</CardGroup>