Connect the PrimeSync Studio plugin.

Before you connect

1. Read the public tool manifest and confirm that the tools match your intended work.
2. Add the endpoint through your client's MCP or connector settings. Choose OAuth when the client asks for authentication.
3. PrimeSync will show the requested permissions separately. Uncheck any scope you do not want to grant.
4. Test with a throwaway strategy first. Never include broker keys, exchange keys, account credentials, or live order instructions.
5. Open Account & activity to see tool-call metadata or revoke a connection at any time.

If a tool fails or keeps asking you to generate

A saved connector can retain an older tool list after the service is updated. Refresh its tools, or reconnect the same MCP URL when your client has no refresh option. The current service exposes five research tools, including primesync_import_backtest, plus the optional memory tool. You do not need to enable memory to build or validate a strategy.

The builder must accept strategyProperties and candidate code. The audit must accept approvedBlueprint and approvedBlueprintHash. If those fields are absent, the client has an outdated definition. A legacy audit can still return an unverified static review; it cannot establish approval or qualify a final candidate.

When a blueprint is ready, the connected AI authors the code and submits it for audit in the same interaction. An authoring contract is not generated code. Compilation, backtests and fresh exports remain external evidence; Studio does not run TradingView.

Omitting the simulation count uses up to 30,000 simulations within the compute budget, with the actual count and any reduction disclosed. Explicit counts are never silently reduced, and trades are never truncated to fit. Supply the original trade count to check dataset completeness. Inspect the current input and output schemas if a client reports an argument error.

Permissions

Alert support

Studio can help your connected AI add Pine Script alert logic to a strategy or indicator using the conditions you approve. Specify which events should trigger an alert, when it should fire, and which message fields you need. Alert setup is available on request; it is not automatically included in every strategy build.

Preserve Existing Logic mode protects existing alerts outside your approved changes. The static audit includes basic checks for matching alert and chart-arrow conditions. These checks do not verify every alert path or confirm live notification delivery.

1. Ask for the alert events, timing and message content you want. Distinguish a confirmed signal from an order fill when defining the trigger.
2. Compile and apply the completed script in TradingView. Create an alert there, select the appropriate script condition, and configure its delivery settings.
3. Test the notification and compare its timing and content with the expected chart event before relying on it.

TradingView runs the alerts and handles their delivery. The MCP does not monitor live prices, create or activate alerts in your TradingView account, or send notifications itself.

Example request: “Add alerts for my approved entry conditions. Include the symbol, timeframe and direction, preserve my strategy rules, and explain how to activate and test the alerts in TradingView.”

Strategy Properties: P0–P4 or Custom

Approve the account and execution model before applying a strategy. These Properties profiles are independent of PrimeSync Max's V1–V12 entry profiles. Studio offers them when requested; it does not insert PrimeSync trading logic into a new strategy.

All five reference profiles use pyramiding 1, calculation on bar close, no recalculation after fills, next-tick order processing, Bar Magnifier off, and limit-fill assumption 0. These are starting assumptions; confirm every override. P0 disables margin checks. P0 and P4 are comparison-only models and omit realistic costs. They do not reproduce prop-firm loss rules or guarantee an earlier equity curve.

1. Confirm the exact symbol/feed, timeframe, session, test dates and account currency. Select a reference profile or supply every Custom value.
2. Approve capital, order sizing and its authority, fees, slippage, margin and execution. Percent of equity is allocation, not risk at the stop. An order's qty can override the default order size.
3. Studio binds the snapshot to the blueprint and provides a Pine declaration for the connected AI to incorporate. An Inputs dropdown cannot rewrite TradingView's Properties tab. Existing code changes remain limited to the scope you approve.
4. Compile the exact code, verify the effective Properties in TradingView, then export fresh trades. Send strategyProperties, propertiesConfirmed and the approved Properties fingerprint with the CSV. A trades CSV alone cannot confirm these settings.
5. Carry the importer's Properties fingerprint into each candidate. Final selection requires confirmed, matching configurations with realistic costs; mixed settings need a comparable rerun. Existing callers without a snapshot can inspect provisional results but cannot complete final selection.

For futures, the emulator's approximate margin requirement is price × point value × quantity × margin percentage / 100. A 10,000 USD P2 account can still be too small for a particular MNQ price. Futures and CFDs can produce different trades because of contract value, feed, session, fees and fill assumptions; increasing capital only addresses funding eligibility.

Example request: “Use P2 for my MNQ strategy. Confirm its Properties and any overrides, preserve my entry logic, and carry the applied settings through the audit, CSV import and Monte Carlo comparison.”

Hosted safety limits

CSV structure and research or Monte Carlo work are bounded before computation. One request may resample at most 25,000,000 aggregate trade observations across every baseline, candidate, and cost-stress path. Oversized requests return a specific safe limit and recovery action rather than silently truncating trades or risking a hosted timeout.

Optional personal memory

The five research tools remain available at no charge and work without memory. To use primesync_user_context, reconnect if necessary and select the optional memory:manage permission. It is unchecked by default. Then explicitly ask Studio to remember the preferences or project summary you choose; granting permission alone does not turn memory on.

For example: “Remember that I prefer Arabic, beginner explanations, and Pine Script.” For a temporary override: “Use Python for this request only.” Say “Continue my strategy” to resume; if several summaries exist, Studio asks which one. Use Account memory controls to inspect, export, reset, or turn off and erase memory. Correction is also available through the context tool.

To continue without memory, carry the exported project's requestContext into the next request. It contains the preferences and portable summary in the input format. Deleting server memory does not delete copies you exported. If supplied baseline code differs from the remembered version, Studio asks you to supply the current source or confirm a temporary branch.

Preferences guide communication and candidate comparison; risk, sessions, and execution still need strategy-specific approval. Custom objective weights express your intent; a tool without that weighted objective asks you to choose a supported metric. Fixed onboarding questions and summary labels are available in English, Arabic and Hebrew. The connected AI translates open-ended details and renders code-comment style without changing tested code. Raw code and complete evidence must still be supplied when required.

Connection limit

Each PrimeSync account can keep up to three active OAuth connections. Revoke an old device or client from Account & activity before adding another.

MCP access

PrimeSync Studio MCP tools have no usage charge and do not consume PrimeSync AI website or widget question credits. Standard connection security and rate limits still apply.