# OptAlgo > OptAlgo is where an AI agent develops, tests and runs trading strategies for its user: a backtest engine that runs the agent's own Python strategy code on real candles and grades how far the result can be trusted, paper trading in the user's My Bots (real market data, no money), and live trading on the user's exchange (Binance, Bybit, OKX, Alpaca) once the user decides in the app. Every plan can connect an agent (Free: backtests and paper only). AI agents (Claude Code, Codex and others): read the agent guide first (start at https://docs.optalgo.com/llm/index.md, or the whole guide in one file at https://docs.optalgo.com/optalgo-llm.md) and follow its path in order. Section 0 connects you to the user's account (a device login the user approves in the app; the key is written to ~/.config/optalgo/credentials.env and never passes through the chat; never ask the user to paste a key). A chat-only assistant (Claude.ai, ChatGPT) without a shell uses the OptAlgo connector instead: backtests and paper bots, no key. Then section 1, a 2-minute paper demo the user sees in My Bots; then section 2, develop a strategy (discover markets, validate for free, backtest, read the honesty grade first, iterate within the plan's quota). Then section 3, deploy a backtested strategy (grade A to C) to run on OptAlgo, on paper, with a free dry run first; section 4, watch its decisions; section 5, request live, which only the user confirms in the app (never claim live until GET /v1/strategies/{id} says live). Every refusal carries `fix` and `next`: follow them. TradingView alerts are optional (section 9), only for users who already have them. ## Agent guide (small files, start here) - [Start here: the index](https://docs.optalgo.com/llm/index.md): the loop (markets, SDK, validate, run, status, result with the honesty grade first, paper or hosted deploy, watch, compare), the safety rules, and every section below with its URL. Under 15 KB - [connect](https://docs.optalgo.com/llm/connect.md): connect an agent (device login, credentials file, claim code, scopes) and the 2-minute paper demo in My Bots - [sdk](https://docs.optalgo.com/llm/sdk.md): write strategy.py with the SDK, several symbols in one strategy (reference series, multi-leg, funding and open interest), the lookahead check, discover markets, validate for free - [backtests](https://docs.optalgo.com/llm/backtests.md): the backtest loop and run spec, baskets, reading a result (honesty grade, search verdict, net of fees), quota, helpers - [research](https://docs.optalgo.com/llm/research.md): avoid overfitting, holdouts, walk-forward research runs - [hosted](https://docs.optalgo.com/llm/hosted.md): run a strategy on OptAlgo on paper, watch it, compare with its backtest, ask to go live - [errors](https://docs.optalgo.com/llm/errors.md): every error code with what to do - [tickets](https://docs.optalgo.com/llm/tickets.md): report problems and missing capabilities automatically, without asking the user - [reference](https://docs.optalgo.com/llm/reference.md): the endpoint reference, recipes, optional TradingView alerts ## Canonical reference - [OptAlgo LLM reference](https://docs.optalgo.com/optalgo-llm.md): the same guide in one Markdown file (about 160 KB, for CLI agents that can read it whole): connect, paper demo in My Bots, develop and backtest a strategy, run it on OptAlgo and go live, errors with fixes, the full API reference, recipes and safety rules ## Docs - [Chat app connector](https://docs.optalgo.com/ai-agents/connectors): use OptAlgo from Claude.ai, ChatGPT or Claude Code without an API key: the remote MCP connector at https://mcp.optalgo.com/mcp (OAuth, approved once in the app), its tools (markets, validate, backtest, paper bots), quota, and why it is paper only on every plan - [API keys](https://docs.optalgo.com/api/api-keys): connect an agent on any plan without handling the key (device login, claim code), plans and backtest quota (Free 50 units a month, Plus 30 and Pro 200 a day), scopes, autonomy and limits, rate limits, logging - [API reference](https://docs.optalgo.com/api/reference): for developers writing their integration once: base URL, versioning and deprecation policy, authentication and the credentials file, error model, rate limits, idempotency, pagination, minimal Python and JavaScript clients - [OpenAPI schema](https://api.optalgo.com/v1/openapi.json): the machine-readable API description (copy at https://docs.optalgo.com/openapi.json; rendered at https://docs.optalgo.com/api-reference.html) - [My Bots overview](https://docs.optalgo.com/my-bots/overview): single-symbol, auto and multi-symbol bots; paper and live; lifecycle - [Multi-symbol bots](https://docs.optalgo.com/my-bots/multi-symbol): one key for many coins; slots, per-coin weights, pausing a coin, plan caps, per-coin closes, conversion - [Partial close](https://docs.optalgo.com/my-bots/partial-close): close_ratio semantics and stop re-arming - [Signal flow](https://docs.optalgo.com/signal-params/signal-flow): webhook endpoint, responses and every check in order - [Troubleshooting](https://docs.optalgo.com/signal-params/troubleshooting): rejection reasons and fixes - [Reliability and safety](https://docs.optalgo.com/platform/reliability): how the platform runs, security, delivery, retries, hedge mode, several bots on one account, position protection and reconciliation ## Optional - [TradingView integration](https://docs.optalgo.com/signal-params/examples/tradingview): alert templates, only for users who already have TradingView alerts - [Signal parameter reference](https://docs.optalgo.com/signal-params/parameters/core-trading): field-by-field reference - [Position types](https://docs.optalgo.com/signal-params/position-types/overview): all position_side values