nomic documentation
AI trading agents in production, under deterministic guardrails they cannot talk their way past.
nomic runs AI trading agents against a venue under risk controls the agent cannot argue with, and keeps a complete, readable trace of every decision and why.
This documentation is being written. The seven sections below are the shape it will take; each one names the issue that fills it and the source its text comes from, so the structure can be reviewed before the content exists.
The sections
- Start here — What nomic is, who it is for, and the shortest path from an account to a run you can read.
- Concepts — The vocabulary: runs, agents, teams, ticks, decisions, wallets, the ledger, theses.
- The risk engine — The deterministic layer between an agent and the market: the rules, the refusals, and why a refusal is a feature.
- Operating a run — Starting, amending, pausing, stopping and liquidating; approvals, notifications and what the auto-stops do.
- Connecting accounts — Provider keys and brokerage credentials: what the platform stores, what it never sees, and what a missing credential does.
- Reference — Every setting, bound and code: the settings reference, the reason codes, the skip reasons and the glossary.
- Workspaces, roles & security — Who can see and do what: workspace roles, platform admission, share links, and how credentials are held.
What is deliberately not here
There is no API reference. The /v1 surface serves this platform's own web tier: it
accepts one credential, a session token, and no third-party client can call it. Publishing it
would document a private contract as though it were a public one.