Agent setup for Scry

Scry is an MCP server at https://mcp.scry.io. Add that URL to your agent and approve the sign-in when it asks; the agent then has schema, sql, embed, feedback and the rest of the roster that tools/list returns, including rerank, memory, vectors, and whoami. The per-client commands for ChatGPT, Claude.ai, Claude Desktop, Claude Code, Codex, and Cursor are on the Connect page (https://scry.io/docs/connect); headless runs pass an API key as a bearer header instead of signing in.

Agents without an MCP client use the HTTP API at https://api.scry.io with SCRY_API_KEY as the bearer credential on every request. Keep the key in ~/.config/scry/env. Every query runs authenticated: under an account key, or per query via x402 payment authorization.

The dashboard at https://scry.io/dashboard has your personal key. From code: create an account with the first request (its response returns a session_token; an inbox that already holds an account signs in with its password instead), then mint a key with the second. Queries run only once the email address is verified: the mail contains a link ending in #verify=<token>, and the same token finishes verification from code with the middle call below, whose reply carries a session_token of its own. The scry product scope is required alongside capability scopes, and the key is shown once: query access needs scry plus read or write, and vector minting needs scry plus embed or write.

curl -X POST https://api.scry.io/v1/auth/signup \
  -H "Content-Type: application/json" \
  -d '{"email":"...","password":"..."}'

curl -X POST https://api.scry.io/v1/auth/signup \
  -H "Content-Type: application/json" \
  -d '{"action":"verify_email","token":"<token from the mailed #verify= link>"}'

curl -X POST https://api.scry.io/v1/auth/api-keys \
  -H "Authorization: Bearer <session_token>" \
  -H "Content-Type: application/json" \
  -d '{"name":"agent","scopes":["scry","read"]}'

Over either surface the agent calls schema before writing SQL. On the MCP schema tool the default mode: "index" is the one-line catalog of relations; mode: "contract" returns the product contract (admission, market, computation, refusals, memory) with the corpus census and live statistics; mode: "full" returns every complete relation block, the discovery authority, larger than most tool-result pipes accept. Over HTTPS, GET /v1/scry/schema takes mode=index or mode=full and refuses any other value, contract included; with no mode it returns the full contracts of the primary-tier doors, the compact index of the depth relations, and the corpus census. The contract itself is GET /v1/scry/context there, and the live statistics are GET /v1/stats.

Install the packaged skill when you want local query patterns and schema notes beside the agent.

npx --yes skills@latest add exopriors/skills