IB Trades History

Fetch trade executions from Interactive Brokers.

IB Connection

TWS or IB Gateway must be running locally with API enabled:

• Paper trading — port 7497
• Live trading — port 7496
• IB_PORT env var — default port when --port is omitted (e.g. IB_PORT=4001 for a Gateway container). Precedence: --port flag > IB_PORT > built-in default. Set it in the shell or a .env file.

Port fallback: If the configured port fails, automatically retry on the other port. If the retry succeeds, save to memory which account type worked (live/paper) and reuse it for all IB skill calls in this and future sessions — until the user explicitly asks for the other account. If both ports fail, ask the user to verify that TWS or IB Gateway is running with API access enabled.

For full trade history beyond ~7 days, the user needs a Flex Web Service token and a pre-configured Trade query in IBKR Account Management.

Instructions

Note: If uv is not installed or pyproject.toml is not found, replace uv run python with python in all commands below.

# Recent trades (last ~7 days via API)
uv run python .claude/skills/ib-trades-history/scripts/trades.py --all-accounts

# Filter by symbol
uv run python .claude/skills/ib-trades-history/scripts/trades.py --all-accounts --symbol AAPL

# Full history via FlexReport
uv run python .claude/skills/ib-trades-history/scripts/trades.py --all-accounts --flex-token YOUR_TOKEN --flex-query-id YOUR_QUERY_ID

# Custom date range (FlexReport)
uv run python .claude/skills/ib-trades-history/scripts/trades.py --all-accounts --flex-token TOKEN --flex-query-id QID --start-date 2025-01-01 --end-date 2025-12-31

# Multiple queries (e.g., one per year to exceed 365-day limit)
uv run python .claude/skills/ib-trades-history/scripts/trades.py --all-accounts --flex-token TOKEN --flex-query-id QID_2025 --flex-query-id QID_2026 --start-date 2025-01-01 --end-date 2026-12-31

# From local FlexReport XML files (no TWS/Gateway needed)
uv run python .claude/skills/ib-trades-history/scripts/trades.py --file trades_2024.xml --file trades_2025.xml --symbol TSLA

# Mix files with date filtering
uv run python .claude/skills/ib-trades-history/scripts/trades.py --file exports/2025.xml --start-date 2025-06-01 --end-date 2025-12-31

# Group option fills into vertical spreads (credit, width, max risk, capture %, RoR)
uv run python .claude/skills/ib-trades-history/scripts/trades.py --file exports/2026.xml --symbol NDX --group-spreads

Arguments

• --port - IB port (default: 7497 for paper trading)
• --account - Specific account ID to filter
• --all-accounts - Fetch trades for all managed accounts
• --symbol - Filter trades by symbol (e.g., AAPL)
• --start-date - Start date in YYYY-MM-DD format (default: Jan 1 of current year)
• --end-date - End date in YYYY-MM-DD format (default: today)
• --flex-token - FlexReport token (enables extended history)
• --flex-query-id - FlexReport query ID (repeatable — pass multiple to merge queries spanning different periods)
• --file - Local FlexReport XML file path (repeatable — pass multiple to merge files). No TWS/Gateway needed
• --group-spreads - Consolidate fills into legs and pair them into vertical spreads. FlexReport sources only

Default behavior (no flags): fetches trades for the first managed account from the live API. Always use --all-accounts unless the user asks for a specific account.

--flex-token / --flex-query-id may be omitted when IB_FLEX_TOKEN / IB_FLEX_QUERY_ID are set in the environment or .env (see env.template). Flags win over env vars.

Data Sources

When using the live API, a data_limitation warning is included in the output.

The live API returns only the current TWS session, despite the "~7 days" the API docs imply — it yields 0 executions on a weekend or for any prior-day lookback, which looks identical to "no trades." Setting ExecutionFilter.time does not widen it. For anything beyond today, use FlexReport. The web service is also aggressively rate-limited ("Statement could not be generated at this time" = throttled, roughly one generation per query per 10–15 min); for large lookbacks prefer a manual XML export via --file.

Spread Grouping (--group-spreads)

Turns raw fills into the trade-level view: a 10-lot order routed across five exchanges is one leg, and its opening and closing legs are one spread.

Requires the openCloseIndicator field, which only FlexReport sources provide. On the live API path the output carries spread_grouping.supported = false with a reason rather than guessing which fills opened versus closed.

Legs are keyed on (account, symbol, expiry, right) — deliberately not trade date, so a spread opened one session and settled the next keeps its closing legs.

Anything that is not an unambiguous two-strike vertical is left in ungrouped_legs with a reason in spread_grouping.warnings, never force-fit into a bogus spread: rolls, ratio spreads, multiple verticals on one expiry/right, single legs, positions still open, and closes whose open predates the window.

spread_grouping.reconciled asserts that spread P&L plus ungrouped P&L equals the raw execution P&L — check it before trusting a report.

Output

Returns JSON with:

• connected - Whether connection succeeded
• source - Data source used (reqExecutionsAsync or FlexReport)
• filters - Applied filters (dates, symbol, account)
• data_limitation - Warning about API date limits (only when using live API)
• execution_count - Total number of executions returned
• executions - List of individual trade executions
• summary - Aggregated stats per symbol (bought, sold, commission, realized P&L)

With --group-spreads, also:

• legs / leg_count - Partial fills consolidated into orders
• spreads / spread_count - Paired verticals, each with type (Bear call / Bull put / Bull call / Bear put), short_strike, long_strike, width, lots, credit, max_credit, max_risk, entry_date/entry_time, exit_date/exit_time, realizedPnL, commission, netPnL, capture_pct, return_on_risk_pct, settled_at_expiry, same_day
• ungrouped_legs - Legs that are not a clean vertical (only when non-empty)
• spread_grouping - supported, warnings, ungrouped_leg_count, and the reconciled / execution_pnl / spread_pnl / ungrouped_pnl check

max_risk assumes a defined-risk vertical: (width − credit) × lots × multiplier. capture_pct is realized P&L as a share of max credit; return_on_risk_pct is realized P&L over max risk. On an iron condor the two wings are separate spreads — summing their max_risk overstates true exposure, since only one side can lose at expiry.

If not connected, explain that TWS/Gateway needs to be running.

Dependencies

• ib-async