Max Agent
Articles

Connecting an exchange bot: API access, monitoring and recovery

A proposed operating workflow for spot automation: approve permissions, keep withdrawals disabled, verify IP restrictions, reconcile test results and define restart conditions.

MaxData EditorialPublished Updated

Write the operating agreement before connecting

Start with a short operating agreement for the connection. Name the account, the person responsible, the intended action and the evidence required before enabling execution. Throughout this article, permission choices, testing steps and recovery actions are our proposed editorial workflow. They are not verified instructions for a particular exchange or a guarantee of safety.

The MaxData bot documentation describes automated purchases and sales triggered through Pro Alerts, Listings or X-GPT. We use that documented relationship between a trigger and an action as the starting point for the agreement. Trading automation documentation.

Write a precise authorization sentence: “This connection may perform this action on these instruments when these conditions are satisfied.” Add a separate list of prohibited actions. In our proposed workflow, withdrawals and activity outside the approved test scope belong on that list.

This article does not verify exchange menus, API compatibility or operator IP addresses. Before proceeding, require an official exchange instruction and the bot operator's connection instructions. Record their links, your verification date and the configuration they cover. Treat an unanswered requirement as a reason to leave execution disabled.

For a beginner, we suggest reviewing one connection at a time. For an experienced operator, use the same agreement for each additional connection and assign a version number. Avoid granting approval to a vague description such as “the usual settings”. Require a written scope that another reviewer could reproduce.

Approve permissions individually

Our proposed permission checklist separates observation from execution. For observation, request read access only if the integration supports that arrangement. For execution, propose adding only the spot trading access required by the approved task. Leave unrelated categories unselected unless the documented task explicitly requires them and you separately approve them.

For every permission visible in the exchange panel, record its exact label, the relevant official explanation and your decision. If the label is unclear, pause configuration and ask the exchange to explain its scope. Do not substitute an interpretation based on what another exchange calls a similar setting.

Our convention for a trading bot key is to keep withdrawal access disabled. If the integration asks you to enable withdrawals, stop this onboarding workflow and obtain an explanation before reconsidering the connection. Do not test the restriction by making an actual withdrawal. Instead, require configuration evidence and an official description of the setting.

We also propose one separately named key per connection, an accountable owner and a review date. Put an inventory reference in the operating agreement rather than the secret itself. Exclude secrets from support questions, shared notes and test reports. These are organizational recommendations; we do not attach a measured protection outcome to them.

Before approval, have the reviewer compare the agreed permission list with the saved configuration. Record discrepancies explicitly. A comment such as “looks fine” is insufficient under our proposed convention because it does not identify what the reviewer accepted.

Proposed checklist

Approve observation and execution separately

Observation
Purpose
Review data and conditions
Proposed access
Read access, if supported
Withdrawals
Keep disabled
Approval evidence
Reviewed observation record
Execution test
Purpose
Verify one approved action
Proposed access
Read access and required spot trading
Withdrawals
Keep disabled
Approval evidence
Action reconciled with exchange evidence
An editorial permission plan to verify against the selected exchange and integration instructions. Observation and execution are compared by purpose, proposed access, withdrawal policy and approval evidence.

Make IP verification an onboarding gate

For this workflow, treat the IP allowlist as a configuration item requiring evidence from both the bot operator and the exchange. Obtain the operator's official connection instructions and check the exchange's official description of its IP restriction controls. We deliberately provide no addresses to copy into a live account.

Our proposed checklist asks four questions: which environment does the supplied list cover, where was it published, who approves changes, and what test follows a change? Record the answers beside the configuration. Require an explanation for an address supplied without an environment or a documented owner.

If you cannot establish the required restriction, mark the gate as incomplete. Our workflow does not prescribe broadening access to make a connection work. It permits returning to an observation stage, when supported, or abandoning that connection.

Plan a separate review for an infrastructure change. Ask the operator how to obtain the replacement list and how to verify the affected connection. Record the new approval independently of the old one. Do not carry a previous approval forward simply because the connection name is unchanged.

Choose the spot scope deliberately

The supplied MaxData bot description identifies spot markets as its trading scope. It also describes selecting individual tokens and sale options covering all held altcoins, either excluding or including BTC. These are statements about the documentation, rather than confirmation of a particular connection's availability. Documented bot scope and actions.

For an initial execution test, our proposed configuration uses an explicitly named instrument and an approved action size. Review broad sale selections separately before allowing them. Specify whether the agreement authorizes purchases, sales or both. Include the account and settlement asset in the written scope.

Add your own total allocation limit, permitted number of actions and test end condition. This article proposes no universal amount or percentage. The operating task here is to document boundaries and compare them with the configuration, without recommending an asset or an expected return.

Ask the reviewer to read the scope back in ordinary language. For example: “The test permits this instrument and this action, and ends at this condition.” If the two reviewers describe different scopes, leave the test unapproved until the wording and configuration agree.

Separate trigger research from connection testing

The English Pro Alerts documentation distinguishes an Alert based on one condition from a Pro Alert combining several conditions. It describes a time window and a sleep period after activation. Pro Alerts definitions.

Our proposed trigger record contains the metric definition, data source, interval, threshold, condition window and repeat behavior. Add an observation horizon and a place for measured delay. If the metric definition or source is unresolved, mark it unresolved rather than treating the chart label as a complete specification.

The English backtest documentation describes checking historical trigger counts and displaying average and median price changes before and after an alert. Backtest output. The supplied Polish backtest material also identifies fees and slippage as possible differences in actual execution. Execution considerations.

In our research worksheet, keep historical observations separate from future expectations. Record the data period, event count, analysis horizon, costs and execution assumptions. Include questions about liquidity, order size and how much parameter selection depended on the same historical sample. Do not use a favorable historical result as approval of API configuration.

For connection testing, first review the condition interpretation. Then use a non-executing test mode if the selected tool provides one. Move to a bounded execution test only after reviewing that result. Define the expected action before the test, and compare it with the exchange evidence afterward.

Reconcile events with a monitoring record

Our proposed monitoring convention gives each event an identifier and records configuration version, data timestamp, trigger timestamp, intended action, exchange response and reviewer conclusion. Use “confirmed” only when the evidence specified in your own protocol is present.

Choose a review interval and an accountable person. For delay measurements, specify the two timestamps being compared and the measurement method. If either timestamp is unavailable, label the result unknown. Do not replace a missing measurement with an estimate presented as a recorded result.

For a discrepancy, use a short diagnostic worksheet. Ask whether the instrument scope matched, whether the configuration version matched, whether authorization was accepted and whether execution evidence is complete. These are proposed investigative questions, not established causes of a failure.

Record alternative explanations alongside the first hypothesis. Set an invalidation condition: what evidence would cause the reviewer to withdraw that explanation? For our workflow, an unresolved discrepancy blocks further execution tests. Approval should identify the evidence reviewed and the exact next action permitted.

Our monitoring convention

Build an evidence trail for each event

  1. 1

    Identify

    Record the event ID, configuration version and available timestamps.

  2. 2

    State intent

    Write the instrument and action approved for this test.

  3. 3

    Collect evidence

    Record the exchange response and mark missing evidence explicitly.

  4. 4

    Review

    Reconcile the result or block further execution pending explanation.

These are proposed recordkeeping steps, without a promised monitoring outcome. Four proposed monitoring steps: identify the event, record the intended action, collect evidence and decide whether the test may continue.

Prepare the stop and restart procedure

Write the emergency procedure before enabling execution. Our proposed stop conditions include an unapproved action, an unexplained configuration change, missing required confirmation or suspected disclosure of a secret. Assign the decision to a named operator and specify a contact route.

The proposed sequence is to suspend new activations, inspect the account directly at the exchange, consider revoking the connection's API access and preserve the event record without secrets. Include a separate review of open orders. Do not treat a recorded “bot stopped” message as completion of the entire procedure.

For restart, require an explanation of the discrepancy, another permission review, renewed approval of the IP configuration and a repeated bounded test. Record who authorized the restart and what they authorized. If the event cannot be reconciled, our workflow keeps execution disabled.

Rehearse the administrative steps using a written scenario. Ask the responsible person to locate the relevant controls and explain the evidence they would preserve. This is our proposed preparation exercise, not a claim that a rehearsal ensures a successful response.

Proposed recovery sequence

Require a fresh decision before restarting

  1. Suspend

    Stop new activations under the prepared operating instructions.

  2. Inspect

    Review the exchange account and open orders directly.

  3. Review access

    Consider revocation and preserve the event record without secrets.

  4. Authorize a retest

    Explain the discrepancy, approve the configuration and repeat a bounded test.

Stages describe our proposed workflow rather than guaranteed response times. The proposed recovery sequence moves through suspension, direct account inspection, access review and a bounded retest.

Use the help hub for documentation questions

The Help Hub presents MaxData Info Bot as an assistant for questions about MaxData and the workspace. MaxData Help Hub.

We suggest asking where to find the connection instructions, how to establish the supported scope and where to obtain the required IP information. Leave API secrets out of the question. The account link below is a product invitation; it does not approve your connection or recommend a trade.

Sources

  1. MaxData Help Hub – Pomoc i przewodniki Mój workspace — agent.maxdata.app; accessed October 9, 2026
  2. Pro Alerts — agent.maxdata.app; accessed October 9, 2026
  3. Backtests — agent.maxdata.app; accessed October 9, 2026
  4. Backtesty Automatyczne Weryfikowanie Strategii — agent.maxdata.app; accessed October 9, 2026
  5. Automatyzacja Tradingu Boty Handlowe — agent.maxdata.app; accessed October 9, 2026

Start working with your own agent

The agent knows this article context and still follows the assigned department checklist.

Create a MaxData account
Max Agent
Exchange bot API permissions, IP checks and recovery