Pair Trading Grok Bot: Meet the PairSync PairBots

PairSync Research Team · Published 2026-11-16 · 7 min

Seven specialist roles, one Bot, your pairs

What if the pair statistics, watchlist, charts and notes you use could be available to the same AI assistant?

PairSync PairBots is our vision for a pair trading Grok Bot with a team of seven specialist roles. The idea is not seven separate Bots: it is one Grok Bot you use, supported by PairSync's nightly statistical analysis and tools designed to help you follow the pairs you choose.

The PairSync connector is available today. The complete guided Bot installation and ordinary-user journey are still being tested. This article explains what you can connect now, the capabilities under development, and how the pieces are intended to work together.

Start with the statistics

PairSync analyses pairs after each market closes and publishes dated statistical output. A compatible AI client can connect to that output through PairSync's Model Context Protocol (MCP) server.

Instead of asking an AI to estimate every relationship from scratch, you can ask it to open the published figures, explain a pair card or screen pairs using bounds you specify. The refit date tells you which published analysis you are looking at.

For example:

Get the latest PairSync refit metadata for US.

Or:

Show me the latest published crypto pairs ordered by Statistical Composite.

For a closer look:

Open the pair card for the first pair in that list. Explain the refit date, content hash and calibration check.

The connector supplies derived statistics, not an intraday price feed. Historical relationships can change; a statistical result is not a forecast or a direction to trade.

One Bot, seven specialist roles

The PairBots design brings these jobs together around a single Bot:

  1. Briefing: summarise the latest published updates and the pairs you follow.
  2. Refit statistics: retrieve the latest published figures and keep their dates clear.
  3. Screening: find pairs within statistical bounds you choose.
  4. Spread analysis: explain a pair's published spread figures and historical context.
  5. Record: help keep a saved statistical snapshot so you can compare it with a later refit.
  6. Methodology: explain what the measures mean and where they come from.
  7. Paper Desk: support practice records that remain clearly marked Simulated.

These are functional roles, not seven independently running AI agents. The connector provides access to the published statistics today; the local tools, instructions and complete guided experience are a separate part of the project and are not yet a verified public installation.

From watchlist to a working view

The PairSync web terminal already provides a place to explore pairs and build watchlists. The Bot project is intended to complement that experience, not replace it.

The goal is to take your own Bot from a conversation to a working view of pairs you have chosen: a watchlist, a pair-by-pair view, dated statistics and a saved statistical frame. The developing PairScreen design separates the nightly fitted figures from current prices, with the source and freshness made visible.

The intended workflow is to open your Bot from a phone or desktop and ask it for the information you need. Exactly which screens, files and alerts are available on each device still needs end-to-end testing. Opening a chart is not the same thing as continuous monitoring.

Boundary alerts and a journal

We are working towards a Bot that can check conditions you specify and bring a boundary crossing to your attention. A boundary is a value you choose to observe against a saved statistical frame; it is not an instruction to trade. Checks depend on their schedule, data source and availability, so they cannot promise to catch every movement.

We also want a straightforward way to record and review trades you place yourself. Development tools already support some user-reported position records and journal notes; the complete guided experience remains to be proven. User-reported records should be labelled as such, not presented as broker-confirmed. Simulated practice records remain separate and marked Simulated.

Your watchlist choices, notes and decisions belong to you. Any actual trade decision is yours and is made through your own broker.

How the technical pieces fit together

The design has three distinct parts:

  • Published analysis: PairSync produces dated pair statistics after market close.
  • Connector: a compatible AI client calls the hosted MCP server to retrieve those statistics. The server address is https://mcp.pair-sync.com/mcp, using Streamable HTTP and supported authentication.
  • Bot tools: the separate tools under development are intended to keep watchlists, snapshots and journal records in the user's Grok Bot workspace and help with scheduled checks. Connecting to MCP does not install these tools.

The Grok Bot computer is hosted in the user's Grok workspace; it is not necessarily the phone or desktop being used to access it. The provider and any connected services process the information needed to run their services. Keep credentials in supported credential fields or a secret store, never in a chat message.

The design favours short tool summaries and combining work into scheduled checks rather than repeatedly sending large data sets through the Bot. In the tested Bot environment, each scheduled run still uses a Bot turn. Costs and usage depend on the provider, schedule and configuration; no particular saving or uninterrupted operation is promised.

When requesting a pair card or a saved statistical snapshot, check the refit date and content hash. A saved snapshot pins one refit; it does not update itself when PairSync publishes another. Keep the saved fit distinct from any later live price observations.

Connect PairSync to your AI

  1. Sign in to the PairSync terminal and find your existing API key under Settings > API Access.
  2. In a compatible client, add the PairSync connector with server URL https://mcp.pair-sync.com/mcp and Streamable HTTP transport. Follow the connection guide for your client.
  3. If the client uses OAuth, start its connection flow and enter your key on PairSync's approval page, not in an OAuth Client ID or Client Secret field. If it supports custom HTTP headers instead, use X-Api-Key as the header name and your key as its value.
  4. Ask the US refit metadata question above. The client should call get_latest_refit_metadata with {"market":"US"} and return a refit date. Saving the connector alone does not confirm access.

One account includes one API key and 100 free lifetime calls shared across REST and MCP. Normal hourly limits apply. A successful connector check does not install the PairBots tools or verify a different app session.

For more detail, see the connector walkthrough, the machine-readable connection guide, the API reference, and PairSync PairBots.


Statistical research only, not investment advice. General information. Does not consider your objectives, financial situation or needs.