Introducing VTX Insights: From Trade-Chain Evidence to Fleet Action

Understand the mechanism behind every result

A PnL figure answers only one question: what happened?

It does not explain whether a bot lost because of its strategy, thin liquidity, timing, sizing, overtrading, execution quality, exit management, fees, or a configuration that changed midway through the period.

VTX Insights lets an agent reconstruct the full path behind a trading campaign:

The settings generation active for each trade
Decision-time market and account conditions
The VTX news and calendar context available to the bot
Model decisions and retained reasoning
Executions, fills, and position changes
Exits, fees, funding, and realized PnL

The agent can then compare losing chains with profitable counterexamples, distinguish repeated problems from ordinary market variance, and identify the mechanism most likely to improve results.

Analysis can cover one campaign, one bot, selected profiles, your complete fleet, named public VTX profiles or wallets, or the whole public platform.

A dedicated trade-chain skill for Codex

The Codex connector includes a VTX bot trade-chain analysis skill. Users do not need to know the available VTX tools or design a multi-step investigation themselves.

The skill guides Codex to:

Identify which settings actually governed the trades being evaluated
Treat recent configuration changes as new generations instead of mixing them with older behavior
Focus first on material trading campaigns rather than spending time processing irrelevant HOLD decisions
Reconstruct decisions, executions, fills, position changes, and exits
Reconcile fees, funding, and PnL before making performance claims
Inspect what the bot knew and how it reasoned at the time
Compare losing behavior with successful counterexamples
Separate strategy judgment from execution, data, or configuration problems
Recommend the smallest useful and reversible experiment
Ask before making consequential settings or trading changes

The skill can retrieve complete requested evidence without imposing artificial population, history, or analysis-duration caps. For efficiency, it starts with the evidence capable of changing the answer and expands when a broader population is material to the question.

Ask the questions that matter to your trading

Review recent fleet performance:

Analyze all my bots over the last 24–48 hours. I changed some settings recently, so judge each configuration only after it became effective. Reconstruct the material trade chains, identify systemic problems, and recommend the smallest useful changes.

Investigate weekend and low-liquidity trading:

Which bots are opening positions during thin weekend or off-hours liquidity? Check whether they recognized the low volume, what evidence caused them to trade anyway, and whether scheduling, a liquidity rule, or a market-universe change would be the best response.

Compare asset classes:

Compare my recent performance in stocks, stock indices, oil, gold, and other commodities. Separate strategy quality from liquidity and execution effects, then recommend which markets appear most suitable for each bot.

Evaluate a new configuration fairly:

I changed this bot’s prompts, sizing, and tradability settings two days ago. Analyze only the trades made under the new generation and compare them with the immediately preceding configuration.

Find execution problems:

Identify trades where the model’s thesis was reasonable but slippage, fees, order timing, position sizing, or exit management materially weakened the result.

Inspect contradictory behavior:

Find decisions where a bot recognized a serious risk but traded anyway. Show what evidence overrode the concern and whether that pattern also appears in profitable trades.

Run a counterfactual:

What would a higher tradability requirement or stricter liquidity rule have changed for new positions and adds? Treat this as a policy replay, not observed live performance.

Compare public strategies:

Compare these public VTX profiles over the last 30 days. Group them by shared settings and policy rather than identity, then explain which mechanisms appear most repeatable.

Prepare an optimization:

Analyze my complete fleet over the last 30 days. Recommend profile-specific settings changes, show me the proposed differences, and create a snapshot before applying anything. Wait for my approval.

Turn an investigation into controlled action

Insights is not limited to producing reports. With the appropriate permissions, the same agent can carry an approved recommendation through the VTX product boundary.

It can:

Inspect the writable settings available for each bot
Preview and compare proposed changes
Create immutable pre-change settings snapshots
Apply different updates across an owned fleet
Verify the saved state of every targeted profile
Restore an earlier snapshot when an experiment should be rolled back
Change schedules and execution modes
Start, stop, restart, or trigger Trader and Assistant bots
Create, clone, rename, archive, restore, or delete profiles
Configure provider and exchange connections through write-only inputs
Place and manage orders, change leverage, cancel orders, and close positions

This supports workflows such as:

Snapshot every bot’s current settings, move the stock-trading bots to market-hours schedules, add a stricter liquidity instruction, verify every saved result, and give me a rollback plan before starting them again.

Fleet operations are designed not to assume today’s account profile limit is permanent. As users manage larger bot swarms, one agent can coordinate heterogeneous changes without turning the task into dozens of repeated UI operations.

Your agent supplies the judgment

VTX provides the trading evidence, deterministic calculations, permission boundaries, and protected action paths. The connected agent chooses how to investigate, calculate, recommend, and respond.

VTX does not grade, approve, correct, or rewrite the model’s conclusions. Models can make mistakes. Users remain free to challenge an assumption, request the supporting evidence, change the methodology, or ask another model to investigate the same question differently.

The agent can also combine VTX tools with its native web research, files, terminal, calculations, and other installed capabilities. Outside research is supplemental: it does not replace the historical VTX context that the bot actually received.

Permissions remain explicit

Analysis, settings, runtime control, profile management, connection updates, and trading are authorized separately. The connected agent receives only the permissions the user approves.

Public trading evidence never grants control over another account. Settings changes, bot operations, profile management, and trading remain limited to profiles owned by the authenticated user.

Provider and exchange secrets are write-only and cannot be retrieved through Insights. Billing, login management, account security, and other account-level administration remain directly managed in VTX.

Connect from the workspace you already use

Codex is the recommended and richest VTX Insights experience. Its connector includes both the general Insights skill and the dedicated trade-chain analysis skill.

VTX Insights also supports Claude Code, Cursor, GitHub Copilot, and Google Antigravity. Each host has a documented remote-MCP connection path, with plugin, marketplace, or registry installation available where currently supported.

The complete public guide—including setup instructions, compatibility status, prompt ideas, capability discovery, and troubleshooting—is available at VTX Insights. Sign-in is required when connecting an agent or accessing private data and authorized actions, not for reading the guide.