Closed beta. Access by invitation: access@slatemark.ai
What you get with Slatemark.
One MCP endpoint your AI client plugs into; three things you get out of it. A senior-analyst skill that knows how to think through a trade. A persistent journal that remembers your theses, your rules, and your patterns. A deep read-only data surface that grounds every claim in a primary source you can cite.
- Read-only by design. Every tool pulls data; none write to your broker.
- Not investment advice. Slatemark is not a registered investment adviser; the methodology and rule parameters described below are user-configurable starting points, not recommendations tailored to your circumstances.
An opinionated methodology the user downloads from Slatemark as a Claude / Anthropic skill (or equivalent for other AI clients): adversarial review on trade pitches (pressure-test before acting), multi-tool decomposition on broader questions, level-grounded TA at the prices the user is actually acting on, and options analysis with greeks / IV rank / structure selection. Every tool returns a primary source and timestamp so the AI can cite numbers rather than recall them from training data.
Plan a trade with thesis, levels, rule references, and catalysts; close with structured exit prices. Then audit yourself: analyze_journal_patterns surfaces where your own win-rate and R-multiple skew across position class, lifecycle, day-of-week of entry, catalyst presence, and stop discipline. Account-profile framing keeps multi-account users in the right risk-capacity bucket.
Decision discipline lives in versioned, editable rules: concentration caps, position lifecycles, hedge management, sizing from risk, tax discipline. Each journal entry pins the exact rules it was sized against; future sessions flag drift when a rule's body changes after fill so you don't act on a stale framework without seeing it.
Slatemark never sends an order, sets an alert, or writes to your broker. Every tool pulls data; your fills and positions stay where they live. No write surface means no blast radius. The journal is the only write path, and the target is your own data.
Real-time quotes, intraday bars (extended hours included), and full option chains with greeks and IV via your linked broker. Without a broker linked, the same tools fall back to delayed market data automatically. Greek-less, but enough to keep an AI conversation moving while you decide whether to link.
Sharpe / Sortino / Calmar, beta vs. any benchmark, realized vol regimes, anchored VWAP, Donchian channels, ATR stops, pair spreads with half-life, support / resistance, mean-reversion score. Composable over any candle series, plus the full TA-Lib indicator set.
Primary-source EDGAR: 10-K / 10-Q / 8-K (and 20-F / 6-K for foreign issuers), Form 4 insider transactions, 13F institutional holdings, and structured company financials. Cited with the actual filing URL, not a summary.
FRED economic releases (CPI, PPI, NFP, PCE, GDP, JOLTS) with upcoming-print dates, the scraped FOMC meeting calendar with SEP flags and press-conference URLs, and Treasury auction results with bid-to-cover and primary-dealer takedown. Know what's pricing in before you trade through it.
Polygon news with per-ticker sentiment, Finnhub earnings calendar and analyst rating distribution, EIA weekly petroleum / gas / power data, CFTC Commitments-of-Traders positioning across commodities and financial futures, and FINRA short interest plus Reg SHO daily short volume.
Broker tokens are minted per account, encrypted at rest with a key unique to this environment, and isolated to your own account. No shared key, no cross-tenant leakage, no token in a config file on someone else's laptop. Schwab today; more brokers on the roadmap.
One MCP endpoint, 9 clients out of the box: Claude Desktop, Claude.ai, Claude Code, ChatGPT, Codex, Gemini, Gemini CLI, Cursor, OpenCode. Mint per-client credentials from the dashboard once; revoke any of them in one click without touching the others.