LadderGraph is the system of record for $25-par baby bonds and preferred shares: full stepped call schedules extracted from SEC prospectuses, verified by a two-model gate, and served to you and your AI with the source on every fact.
Get free accessConnect your AIMost services store one call date at par. Real securities step down: $25.50, then $25.25, then par, and each step changes the yield math. LadderGraph stores the whole ladder with a prospectus source per step, and computes yield to worst across every future step plus maturity.
One MCP endpoint, connected in a click. Your assistant answers with the rule that fired and the prospectus behind it, never a vibe.
Agents call read-only tools; the graph answers in structured JSON with provenance and rule traces, so an agent can show its work without a second lookup.
In Claude: Settings, Connectors, add custom connector with the MCP URL. In ChatGPT: add it as an app. One URL either way.
Sign in once. Your assistant gets read-only tools: search, ladders, portfolio status, propose_roll. It can never trade or write.
"Is anything in my sleeve near a call?" The answer arrives with the rule that fired and the prospectus it came from.
# or straight from the API curl -s https://api.laddergraph.com/v1/securities/SAX/ladder {"data":[{"effective_date":"2028-08-26","redemption_price":25, "kind":"optional_call","verified":1,"source_url":"https://www.sec.gov/..."}],"count":1}
Structural facts come from the prospectus, pass an extract-then-critic gate across two independent models, and carry their source forever. Live values are timestamped. Anything unverifiable stays blank and flagged, never back-calculated. Rule-based analysis over cited data, not investment advice.