# Cascadian > Cascadian is the agent-native causal decision substrate — the layer an AI agent > calls before money moves. It is one MCP server in front of one Postgres, and it > hands an agent five gifts between a question and a decision: tritemporal memory > (every fact carries as_of and available_at, so the system physically cannot lie > about what was knowable), a governed causal graph (humans own truth; agents only > propose), an identification gate (no causal estimate without a valid > identification basis — otherwise exploratory, and labeled), a knock-on simulator > (simulate_intervention: do(X=x) on the approved graph, every downstream effect > with its assumptions), and a decision ledger (expectations frozen before > outcomes, so judgment compounds). Six invariants are enforced structurally — by > schema constraints, append-only storage, and mandatory service paths — never by > prompt. ## Connect over MCP claude mcp add cascadian https://mcp.cascadian.ai/alpha ## Pages - Home: https://cascadian.ai/ — The substrate story, the gap it prices, the architecture, pricing - How it works: https://cascadian.ai/how-it-works — From a question to a decision that compounds - Causal intelligence: https://cascadian.ai/causal-intelligence — Why correlation fails when the stakes are highest - Architecture: https://cascadian.ai/architecture — One MCP server. One Postgres. Six invariants - Concepts: https://cascadian.ai/concepts — The vocabulary of causal forecasting - Who it's for: https://cascadian.ai/who-its-for — Agents that act — and the desks that supervise them - Proof: https://cascadian.ai/proof — The public, scored decision ledger - Signals: https://cascadian.ai/signals — The world, priced in real time — live market odds, attributed and timestamped - Forecasts: https://cascadian.ai/forecast — Forecast exemplars — real market readings with the causal layer's illustrative overlay - Indicators: https://cascadian.ai/indicators — The signals the substrate watches - Services: https://cascadian.ai/services — Hands-on AI strategy & engineering - Investors: https://cascadian.ai/investors — The round, the deck, the founders - Brand guide: https://cascadian.ai/brand — The Trace design system - Changelog: https://cascadian.ai/changelog — What shipped (from the real build history) and the build-plan roadmap, labeled not shipped - Contact: https://cascadian.ai/contact — Request a briefing or alpha access - AGENTS.md: https://cascadian.ai/AGENTS.md — Connection notes for agent visitors ## Machine-readable endpoints - Context bundle: https://cascadian.ai/api/context — One GET: positioning, live market rows (attributed, timestamped), the illustrative causal pathway, connection info. JSON, CORS open, revalidated every 5 minutes - Tool manifest: https://cascadian.ai/api/tools.json — The designed MCP tool surface (~27 tools, four groups) plus the six invariants, machine-readable. Alpha is MCP-only; REST paths are planned - OKF bundle: https://cascadian.ai/okf/index.md — Open Knowledge Format (Google v0.1) — Cascadian's core concepts as cross-linked agent-readable markdown. Start at /okf/index.md - llms.txt: https://cascadian.ai/llms.txt — This file ## MCP & API documentation ### Get Started - Overview: https://cascadian.ai/mcp-api - Quickstart: https://cascadian.ai/mcp-api/quickstart - Authentication & Approvals: https://cascadian.ai/mcp-api/authentication - The Decision Object: https://cascadian.ai/mcp-api/forecast-object - Conventions: https://cascadian.ai/mcp-api/conventions ### Concepts - The Four Memory Networks: https://cascadian.ai/mcp-api/concepts/memory-networks - The Governed Causal Graph: https://cascadian.ai/mcp-api/concepts/governed-graph - The Knock-on Simulator: https://cascadian.ai/mcp-api/concepts/knock-on-simulator - The Decision Ledger: https://cascadian.ai/mcp-api/concepts/decision-ledger ### Memory - Bootstrap: https://cascadian.ai/mcp-api/tools/bootstrap - Ingest Source: https://cascadian.ai/mcp-api/tools/ingest_source - Assert Fact: https://cascadian.ai/mcp-api/tools/assert_fact - Propose Memory: https://cascadian.ai/mcp-api/tools/propose_memory - Link Memories: https://cascadian.ai/mcp-api/tools/link_memories - Supersede Memory: https://cascadian.ai/mcp-api/tools/supersede_memory - Search Memory: https://cascadian.ai/mcp-api/tools/search_memory - Get Facts: https://cascadian.ai/mcp-api/tools/get_facts - Get Entity: https://cascadian.ai/mcp-api/tools/get_entity - Get Memory: https://cascadian.ai/mcp-api/tools/get_memory - Context Pack: https://cascadian.ai/mcp-api/tools/get_context_pack - List Contradictions: https://cascadian.ai/mcp-api/tools/list_contradictions ### Causal - Propose Nodes: https://cascadian.ai/mcp-api/tools/propose_nodes - Propose Edge: https://cascadian.ai/mcp-api/tools/propose_edge - Get Graph: https://cascadian.ai/mcp-api/tools/get_graph - Validate Graph: https://cascadian.ai/mcp-api/tools/validate_graph - Identify: https://cascadian.ai/mcp-api/tools/identify - Estimate: https://cascadian.ai/mcp-api/tools/estimate - Simulate Intervention: https://cascadian.ai/mcp-api/tools/simulate_intervention - Attribute Change: https://cascadian.ai/mcp-api/tools/attribute_change ### Decisions - Open Decision: https://cascadian.ai/mcp-api/tools/open_decision - Record Recommendation: https://cascadian.ai/mcp-api/tools/record_recommendation - Record Decision: https://cascadian.ai/mcp-api/tools/record_decision - Record Outcome: https://cascadian.ai/mcp-api/tools/record_outcome - Propose Learning: https://cascadian.ai/mcp-api/tools/propose_learning - Review Calibration: https://cascadian.ai/mcp-api/tools/review_calibration - Get Decision: https://cascadian.ai/mcp-api/tools/get_decision ### Governance - Request Approval: https://cascadian.ai/mcp-api/tools/request_approval - Pending Approvals: https://cascadian.ai/mcp-api/tools/get_pending_approvals - Reconstruct Decision: https://cascadian.ai/mcp-api/tools/reconstruct_decision - Detect Leakage: https://cascadian.ai/mcp-api/tools/detect_leakage - Capabilities: https://cascadian.ai/mcp-api/tools/capabilities ### Reference - Errors & Limits: https://cascadian.ai/mcp-api/errors - Access & Pricing: https://cascadian.ai/mcp-api/pricing ## Contact briefings@cascadian.ai ## Status Pre-launch alpha. Access is gated while we run paid pilots. Market readings on /signals and /forecast are REAL — attributed to their source (Polymarket/Kalshi) and stamped with a fetch time. Cascadian's causal overlays, example tool responses, and ledger figures remain illustrative and carry a visible ILLUSTRATIVE/SAMPLE label.