> For the complete documentation index, see [llms.txt](https://docs.visionmakers.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.visionmakers.io/platform/intelligence-layer.md).

# Intelligence Layer

The AI co-pilot that prices the route, explains the choice and arms conditional orders

The Intelligence Layer adds reasoning to execution. It prices the route, explains why it chose that
route, interprets risk, and can arm orders that execute when conditions you set are met.

## What it adds over vDEX

| Capability | Behaviour |
| --- | --- |
| **Route pricing** | Compares paths on output after slippage, gas and fill probability, then states which won and by how much |
| **Risk interpretation** | Liquidity, concentration, contract flags and deployer behaviour, with components exposed |
| **Conditional orders** | Price triggers, armed and cancellable |
| **Scheduled agents** | Recurring workflows that pre-stage trades |
| **Plain-language restatement** | What will happen, in words, before you sign |

## How an instruction resolves

Take *"Buy 5 SOL if it drops under $120."*

1. **Parse.** Intent `conditional_order`, asset `SOL`, size `5`, trigger `< 120 USD`.
2. **Ground.** Does SOL exist on a supported network, is there routable liquidity at that size, is your balance sufficient.
3. **Reason.** Form a judgement and assemble the evidence behind it.
4. **Explain.** Render at your demonstrated level of expertise.
5. **Offer.** Present one confirmable control.
6. **Sign.** You approve. Only then does anything execute.

The order then appears as:

> **Buy 5 SOL · trigger at $120 · Armed**
> SOL now $127.40. Fills on Solana.

## Conditional orders

| Property | Detail |
| --- | --- |
| Trigger types | Price thresholds, in either direction |
| Visibility | Armed orders stay listed for their whole life |
| Cancellation | Any time before the trigger fires |
| Restatement | Fill conditions restated in plain language at arming |
| Execution | From your own wallet, within the rules you set |

<div class="callout">
<span class="callout-label">Armed is not executed</span>
The assistant arms orders; you fire them. A condition can still be met in circumstances you did
not anticipate, for example a flash move that reverses within the same block. Set conservative
limits and review armed orders regularly.
</div>

## Risk interpretation

Risk is assessed continuously per token across five dimensions:

| Dimension | What is measured |
| --- | --- |
| **Liquidity** | Depth, stability, and whether it grew with price or lagged it |
| **Holder concentration** | Top-10 share, single-wallet dominance, distribution shape |
| **Contract flags** | Mint authority, transfer restrictions, upgradeability, honeypot patterns |
| **Deployer behaviour** | Sells, transfers and timing since deployment |
| **Flow anomalies** | Deviation from the token's own established pattern |

A result reads as a judgement with its inputs visible:

> **Healthy liquidity, no risk flags.**
> Liquidity $4.2M · Top 10 holders 18% · Risk flags none

### Why there is no single safety score

A composite number invites false confidence and cannot be argued with. The platform exposes the
components so you can disagree with the conclusion while still seeing what produced it. An
18% top-10 concentration is unremarkable for a mid-cap and alarming for a day-old launch, and
no single number carries that distinction.

## Scheduled agents

Agents chain the above into recurring workflows:

> *"Every morning, brief me on Base and pre-stage anything that clears my rules."*

> **Daily agent · Base morning brief**
> Every weekday at 08:00, delivered in Telegram.
> Discovery, sentiment shift, smart-money flow.
> Pre-staged swaps, nothing moves without approval.

Every agent run is logged against your account: what ran, what it found, what it pre-staged and
what you signed. See [Security](/whitepaper/security).

## Grounding, and what it rules out

The assistant does not answer from model memory. Every factual claim traces to a record in the
intelligence core with lineage attached, and where data is missing or stale the answer says so
rather than interpolating.

The failure mode of a grounded system is *"I do not have that"*, which is recoverable. The
failure mode of an ungrounded one is a plausible fabrication, which is not.

## Limits

**AI output can be wrong.** Grounding reduces fabrication; it does not eliminate error. A
confident, well-evidenced and incorrect interpretation is possible.

**Risk flags are indicative.** The absence of a flag is not a guarantee of safety. New attack
patterns appear faster than detection for them.

**Conditional orders are not stop-losses.** They execute on a condition, subject to liquidity at
that moment. In a gap or a liquidity collapse the fill can be far from the trigger.

See [AI Transparency](/legal/ai-transparency) and
[Risk Disclosures](/legal/risk-disclosures).

## Related

[vDEX](/platform/vdex) · [Mindshare](/platform/mindshare) ·
[Whitepaper section 6](/whitepaper/intelligence)