01
Query it over MCP
Connect to /api/mcp to retrieve supported decision objects, evidence and analytical results through read-only tools. Public reads and the metered packet have separate access terms; the MCP guide lists them.
API & MCP · for the builder of any bot at any door
Read the desk the way a person does, over MCP and the API.
Every door on this site assumes a bot may stand at it: a trading bot, a fund's research agent, an AI founder raising by market. This page is where its builder starts. Retrieve the conditional comparison, its stated objective and the available market evidence, research and calls as structured context, with source times, missing inputs and separate findings, so an agent can tell a market implication from a passage forecast, an execution check or an observed outcome. Coverage differs by decision and tool. As of 1 November 2026 the record names the agent decider and the agent trader as classes and counts zero agent deciders and zero agent traders; the first to appear changes the count here, under a release marked not backward compatible.
What the desk serves a machine
02
Fetch it as plain JSON
Read decoded decisions at /data/decisions.json, with retained analysis where available. Retrieve supported public objects by exact key at /api/objects, or use the published instrument and ontology definitions. The API guide lists coverage and access requirements.
03
Get provenance with the value
Read each instrument’s identity, method version, evidence cutoff and coverage alongside its result. Interpretations reference supporting results and retain their limitations. Historical review records keep their original contracts.
04
Read the desk's calls with their scores
The forecast publication carries each open window's pass or fail probability and the record of resolved windows against their stored base rates, so an agent that acts on a call can weigh it by the score it has earned.
05
Pay for the metered packet
The analytical packet for open MetaDAO windows, including replay inputs and flip-cost calculations, is priced per pull over x402, the HTTP 402 handshake: ten pulls a day free, then $0.02 in USDC on Base or Solana, once metering is switched on; until then the packet does not serve. An unpaid call past the allowance answers with machine-readable terms and how many windows are open, so nothing is paid for an empty tape.
06
Reason over an explicit ontology
The ontology defines decisions, options, execution, measured outcomes and analytical instruments. Its versioned OWL 2 axioms and SHACL shapes are served as MCP resources, so an agent can inspect the meaning and relationships behind a result.
07
Train and evaluate on dated decisions
The dated export gives every row the time Eidos learned it, with resolved outcomes, so a field enters a model's state only once its timestamp is at or before decision time. Researchers take it free under a named citation; a forecasting company's use is scoped and quoted. The pilot page states both.
What one decision context carries
8 fields, each answered today by the read-only tools and keyless URLs named beneath it. The composed call resolves exact proposal identities into public evidence revisions, observation clocks, and explicit gaps. Richer instruments remain available through the tools below.
get_decision_context(proposal)proposal_stateProposal stateWhere the proposal sits in its lifecycle — drafted, funded, recording, settling, finalized — decoded from confirmed program accounts rather than from a scraped page.
get_live_proposals list_replays /api/live-proposalsmarket_stateMarket stateThe conditional books: pass and fail time-weighted accumulators, spot quotes around them, and the depth the print is standing on.
read_decision_market get_live_proposals /api/live-proposals /api/organization-quotestwap_thresholdTWAP and thresholdThe DAO account's configured decision threshold and the current pass edge against it, with the recording window's remaining stretch — the distance to a verdict, not a prediction of one.
get_live_proposals read_decision_market /api/live-proposalstreasury_exposureTreasury exposureThe organization's cited balances and vault history read against what the proposal is asking for, with the fields a source does not cover left visibly empty.
/api/dao-vaults /data/history.jsonevidence_completenessEvidence completenessWhich required fields are present, which are missing, and how current the packet is — the gate every classification has to clear before it is allowed to run.
qualify_evidence /api/proposal-evidence /data/operational-evidence.jsonadversarial_findingsAdversarial findingsEvery monitored attack vector run against the record, each reporting its own indicator count and scope note, plus the mechanism components nothing here defends.
scan_attack_vectors map_attack_surface map_live_exposure /vectors/library.json /data/lane-findings.jsonhistorical_contextHistorical contextThe published cases this record resembles, the standing checks they produced, and the recorded series behind them. The check carries forward; the old conclusion does not.
carry_forward list_replays /data/proposal-window-replay.json /data/market-analytics.jsonprovenanceProvenanceReview, contract version, instrument version, source, and read time on every value — so a caller can tell an observation from an estimate and a withheld review from a zero.
enforce_scope get_review_contracts /ontology/eidos.ttl /ontology/shapes.ttlThe doors a bot may stand at
Add your organization's side of the record.
Every organization is on the record whether it answers or not. Beside what the chain shows, state what it cannot: what a decision is for, who it binds, what you committed to, and the documents behind it, checked and dated.