StrategyFactory.ai

Showing 25 of 49 from the live probe.

• Login

  Check authentication status. If already authenticated, returns your user info. If not authenticated, provides a browser login URL. The user opens the URL, signs in with Google via Clerk, and the page displays a pk_… API key. They paste the key back here and you call authenticate({ key }) to store it for this session.

• Authenticate

  Store an API key for this session. Validates the key against the API and logs you in. The key is held in memory for this session only — never written to disk.

• Create Api Key

  Create an API key for a TraderDev account in one step. Provide your email and password. If no account exists for that email, one is registered automatically. The generated pk_… key is stored for this session automatically — no need to call authenticate() afterwards. Save the key somewhere safe — it is shown only once.

• Whoami

  Verify auth and return the current user's basic info. Recommended as the first call in every conversation to confirm the API key is valid and give Claude context about who it's acting on behalf of.

• Get Pine Codegen Rules

  REQUIRED before writing any Pine for quick_backtest. Returns the full MCP RULE (mcprule.txt) plus the list of 65 parity-tested ta.* indicator names and proven strategy templates. You MUST generate //@version=6 Pine that only uses allowlisted indicators and strategy.exit / entry / close. Broker header MUST use commission=0, percent_of_equity=100, margin_long/short=100. Never emit martingale, strategy.cancel, request.security, arrays, pyramiding>1, or fixed qty=1 on BTC. When the user asks to crea

• Plan Backtest Window

  Pre-flight: resolve a user-requested symbol/timeframe/from/to against live ClickHouse coverage (same data as /market/coverage). Returns the APPLIED symbol and clamped dates plus adjustment reasons. Delisted / missing symbols (e.g. TONUSDT) are remapped to BTCUSDT. Dates outside the archive are clamped. ALWAYS call this (or trust quick_backtest parityAdjustments) so you do not invent unavailable history. When the user asks to create or backtest a strategy, show them the Strategy Tester tip: 100%

• List Strategies

  List the user's saved strategies. Filter by mode: 'dev' = drafts under development, 'deployed' = live strategies. Pine source is omitted from the list to keep responses concise; use get_strategy to read the full source of a specific strategy.

• Get Strategy

  Read a single strategy by ID, including its full Pine source code.

• Create Strategy

  Create a new strategy. The strategy is saved in 'dev' mode by default. Provide a valid Pine Script v5 source. All crypto symbols resolve to Bybit USDT linear perpetual. Accepts fuzzy input: btc, BTCUSDT, BYBIT:BTCUSDT.P, etc. Forex pairs (EURUSD) pass through unchanged.

• Update Strategy

  Update an existing strategy. Only the fields you provide are changed; all others remain unchanged. GIT-COMMIT SEMANTICS — when `pineSource` is in the patch and differs from the current row, the server creates a NEW VERSION instead of mutating in place. The old row stays as read-only history. The response is the new versioned row (status 201) — capture its `id` to keep iterating against the new tip. Trivial edits (name, capital, status, warmupBars, sinks, messageTemplate) update in-place.

• Delete Strategy

  Permanently delete a strategy. This cannot be undone.

• Promote Strategy

  Promote a strategy from 'dev' mode to 'deployed'. Within ~30 seconds the live alerts runtime picks it up and starts evaluating it on every new bar. Fires real Telegram alerts. Returns the updated strategy row.

• Demote Strategy

  Demote a strategy from 'deployed' back to 'dev'. The live alerts runtime stops evaluating it within ~30 seconds. No alerts will fire after demotion. Returns the updated strategy row.

• Search Strategies

  Search public strategies by performance criteria. Use this to find candidates the user can fork and iterate on (the leaderboard equivalent of an API call). All filters are optional and AND-ed. Numeric filters use min/max bands; ratio fields like Sharpe accept negative numbers. The default sort is `profit` (best net P&L%); other axes: `sharpe`, `sortino`, `drawdown` (ascending — lowest drawdown first), `winrate`, `trades`, `recent`. Each result includes the strategy id (which fork_strategy / qu

• Fork Strategy

  Fork a public strategy into your account so you can iterate on it. Returns the new strategy row including its `pineSource`, `id` (the new strategyId), and `version`. WORKFLOW after forking: 1. Read the returned `pineSource`. 2. Modify it — add indicators, tighten filters, change exits, swap thresholds. 3. Call `quick_backtest` with `strategyId=<new id>` and the modified `pineSource`. That commits a new VERSION (git-style) and runs the backtest in one shot. Each subsequent tweak is another

• Pause Strategy

  Pause a deployed strategy. The strategy remains in 'deployed' mode but alerts stop firing immediately. Use this to temporarily silence a strategy without demoting it. Resume with resume_strategy.

• Resume Strategy

  Resume a paused strategy. Alerts will start firing again on the next bar evaluation (~30s). Returns the updated strategy row.

• Search Perps

  Search the Bybit USDT linear perpetual catalog (~570 instruments). Use when the user names a coin ambiguously (e.g. "bonk", "pepe") before create_strategy or quick_backtest.

• Run Backtest

  Re-run a saved strategy on the SAME tv_jul26 parity engine as quick_backtest (not legacy PineTS). Returns a completed result synchronously (resultId + KPIs). Prefer quick_backtest when you already have the Pine source.

• Get Backtest Result

  Get the status and (when complete) the full result of a backtest job. By default this waits up to 60 seconds for the job to finish — so you can call it immediately after run_backtest and Claude will get the result in one step. Set waitForCompletion=false to get the current status without waiting. ⚠️ The response may include a `warnings` array. Same rules as `quick_backtest` — if a `cascade_exit_pattern_severe` warning fires, the result is unreliable and you MUST retry with a corrected strategy

• Quick Backtest

  Run a backtest synchronously on the TV_ENGINE_JUL_26 parity path (Pine Script //@version=6 ONLY). BEFORE calling this tool you MUST call `get_pine_codegen_rules` and generate Pine that obeys mcprule.txt (allowlisted ta.*, strategy.entry/exit/close, pyramiding=1, process_orders_on_close=true). Forbidden: martingale, strategy.cancel, request.security, arrays, pyramiding>1. Match TradingView: 100% equity, margin 100/100, pyramiding 1, commission 0.05%, bar close. ALWAYS use this tool for //@versi

• Get Trades

  Get the full per-trade list for a completed backtest. Accepts either a saved-strategy job ID (from run_backtest) or an adhoc result ID (from quick_backtest — pass `result.id` or top-level `resultId`, NOT `result.jobId`). Each row includes entry/exit prices, quantity, net P&L (commission-inclusive, matching TradingView's "Net P&L" column), gross P&L, commission paid, run-up, drawdown, and cumulative P&L.

• Get Equity Curve

  Get the per-bar equity curve for a completed backtest. Accepts either a saved-strategy job ID (from run_backtest) or an adhoc result ID (from quick_backtest — pass `result.id` or top-level `resultId`, NOT `result.jobId`). Returns barIndex, barTime (Unix ms), equity, drawdown, and netProfit at each bar. Large results are downsampled to maxPoints.

• Parse Strategy Inputs

  Extract all input.* declarations from a Pine Script source. Returns each input's name, type, default value, and bounds (min/max/step for numerics, options for string enums). Use this to discover what parameters a strategy has before running an optimization sweep.

• Optimize Strategy

  Run a parameter sweep over a strategy and return the top N combinations by a chosen metric. Provide either strategyId (loads saved Pine source) or pineSource directly. paramRanges defines which inputs to vary — use explicit values[] or a min/max/step range. The sweep runs up to maxRuns combinations (default 100) in parallel batches of 5. Returns totalRuns, duration, and the top N results sorted by the chosen objective. Note: maxDrawdownPct is minimized; all other objectives are maximized by defa