START HERE — single entry point that returns either a global market snapshot or a thesis-specific context bundle. Global mode (no args): top mispriced edges, 24h price movers, highlights, traditional markets. Thesis mode (thesisId + apiKey): adds causal tree, signal log, and evaluation history for that thesis. Read-only, no rate limit. Use this first; only call get_edges / get_changes / get_world_state if you need that single slice in isolation.

Calibrated world model: ~9,700 live prediction markets distilled into ~800 tokens of real-money probabilities across geopolitics, economics, tech, and policy. Read-only, no auth, no rate limit. Use when you need a compact snapshot of "what the market believes right now"; use get_changes for deltas only, or get_context for the broader bundle including edges and movers.

Incremental diff of the world model since a given timestamp — only the markets whose probability moved. ~30-50 tokens vs ~800 for the full state from get_world_state. Read-only, no auth. Use this for cheap polling loops; use get_world_state for an absolute snapshot.

List live prediction market contracts with current YES/NO prices, 24h volume, and metadata. Read-only, no auth. Use for deep dives on a specific topic; use search_markets if you have a keyword instead of a topic, or get_context for a high-level overview.

Full-text search prediction market contracts by keyword across question text and resolution criteria. Read-only, no auth. Use when you have a free-form term ("OPEC", "Powell", "TSMC"); use get_markets if you only want to filter by predefined topic.

What moved in the last 24 hours: incremental delta of probability changes across all tracked prediction markets, returned as compact markdown (~30-50 tokens). Read-only, no auth, no parameters. Thin wrapper over get_world_delta with since="24h"; use get_world_delta directly if you need a different lookback window or JSON output, or get_context for movers bundled with edges and highlights.

Current mispricings (edges) detected across all public theses — contracts where the platform's causal model disagrees with market price. Read-only, no auth, no parameters. Returns only the edge list; use get_context for edges bundled with movers, highlights, and world state.

AI-generated trade ideas derived from active theses and current market data, each with rationale and target contract. Read-only, no auth, no parameters. Use when you want pre-packaged actionable suggestions; use get_edges for raw mispricings without commentary.

Cross-reference arbitrary text against live prediction markets: paste an article or note, get back the markets relevant to its claims plus an LLM digest. POSTs content to the server; no auth required, no persistence. Use for one-off article enrichment; use monitor_the_situation for scheduled URL scraping with webhook delivery.

List every thesis owned by the authenticated user, with id, title, status, and last evaluation timestamp. Read-only. Requires SF API key. Use to discover thesisId values needed by get_context, trigger_evaluation, inject_signal, and fork_thesis.

Append an external observation (news headline, user note, data point) to a thesis. Stored in the thesis signal log and consumed on the next evaluation cycle — does NOT trigger evaluation by itself; call trigger_evaluation afterward if you need an immediate update. Writes state. Requires SF API key.

Force an immediate thesis evaluation: consumes pending signals, re-scans edges, and updates confidence scores. Side-effectful and LLM-billed (typically 5-30s, may be rate-limited per plan). Requires SF API key. Use after inject_signal when you need fresh output now; otherwise theses re-evaluate on their own schedule.

Create a new thesis from a natural-language statement. The platform parses it, builds a causal tree, and schedules recurring evaluation. Side-effectful. Requires SF API key. Use fork_thesis instead if you want to start from an existing public thesis.

Clone a public thesis (by id or slug) into the authenticated user's account, copying its causal tree as a starting point. Side-effectful. Requires SF API key. Use when you want to iterate on someone else's thesis; use create_thesis to start from scratch.

End-to-end pipeline: scrape one or more URLs (or run a search/crawl/map/extract), optionally analyze with an LLM against a prompt + JSON schema, optionally cross-reference with prediction markets, and return the bundle. Side-effectful (calls Firecrawl + LLM, billed). Requires SF API key. Use for scheduled or one-shot URL ingestion; use enrich_content if you already have the text in hand.

Fetch real-time and historical market data from Databento (CME futures, US equities, crypto). Read-only, billed per Databento usage. Use for OHLCV / trades / quotes on traditional instruments; use get_markets or search_markets for prediction-market contracts instead.