What works with Evrant, and what you own.

Strategy code

A strategy is a C# class with one method, OnCompletedBar, which the engine calls once per completed bar. It receives the bar, session details, the instrument's dated specification, your portfolio and engine-computed indicator values. It returns order intents and chart annotations.

Code runs in a sandbox and must stay inside a versioned allow-list. Ordinary C# is fine: maths, strings, collections, LINQ, records, pattern matching. File and network access, clocks, unseeded randomness, threads, reflection and native code are rejected, partly for safety and partly so that the same inputs always give the same result. The SDK provides a seeded random source and session-time helpers instead. The build records the allow-list version it passed.

The agent can write the code from a recorded specification, or you can write or edit it yourself. Either way it goes through the same compiler, analyser and verifier.

Bringing an existing strategy

There is no converter from Pine Script, NinjaScript, MQL, Python or other platforms. Two routes work:

• Describe the rules in plain English. Evrant keeps your words verbatim, the agent writes a specification you can check, then builds C# against it.
• Rewrite the logic in C# against the SDK yourself.

Expect results to differ from another platform even when the logic matches. Fill rules, session handling and contract rolls differ between simulators. Evrant’s are written down in the one-minute bars guide, and the run records which fill model version it used.

Market data

Evrant uses one-minute OHLCV bars. Strategies can run on that or on longer timeframes, from 5 minutes to 1 day, built from complete one-minute bars. Which markets and date ranges are available is listed in the app's data catalogue.

Market data is licensed from third parties for your research inside Evrant. You cannot bulk-export or redistribute it (see the terms). Uploading your own data is not supported.

Agents and MCP

Any MCP client can connect to Evrant over streamable HTTP with a personal token from your settings. The current catalogue has 101 tools under the scopes read, research, build, compute, mandate, validation. They are the same commands the in-app agent uses, with the same budgets and split enforcement. See them in use on the agent page.

Some decisions stay with people. You can let a token approve a specification or a validation plan if you choose. Sealed-data evaluation can never be delegated: it is always your own confirmation in the app. Why that matters is covered in the validation guide.

What you own

You own the ideas, strategies, code, research records and results you create or upload. We use them only to run, secure and support the service for you, which includes sending them to the AI model providers behind the research agent. We do not share them with other users. The Evrant software, engine and SDK remain ours. The terms are the binding version.

You can read the exact C# of every strategy version inside Evrant, along with its hash, and every result links to the run, data version and settings that produced it.

Leaving Evrant

After your account closes you can ask for an export of your content within 30 days, and you can ask for your personal data in a portable format under the privacy notice. Self-serve downloads of strategies and evidence reports are not available; ask us for an export instead. Licensed market data is not included in any export.

Running an Evrant strategy outside Evrant is not supported. The SDK is part of the Evrant engine.

Trading

Evrant is research software. It tests strategies on historical data and helps you decide whether one has earned a candidate decision. Paper trading and broker connections are not included, and we will not describe them until they exist.

Evrant is research software, not investment advice. Historical and simulated results do not predict future returns. Trading futures carries a substantial risk of loss. Product mockups use illustrative figures.