Tross

INSTALL WITH AI

Set up Tross with your AI assistant

Paste one prompt into Claude Code, Cursor, Copilot or any coding agent. It installs the client for your language, wires the API key safely, and proves the setup with a real call. Every operation page also has its own Install with AI prompt, scoped to that one endpoint and its fields.

Copy this prompt

Set up Tross in this project so it can call EHR and payer-portal operations through the Tross API.
Docs: https://app.ontross.com/docs — machine-readable index: https://app.ontross.com/llms.txt

Rules
- Never ask me to paste my Tross API key into this chat, and never print, cat, grep or echo it. It lives only in .env as TROSS_API_KEY.
- Use only the providers, operations and fields that appear on the Tross docs pages. Do not invent endpoints.
- Make the smallest change to this project's existing configuration: do not switch module systems, build tools or runtimes unless the install cannot work otherwise, and say so if you must.

Steps
1. Tools: if you support MCP, add the Tross sandbox server so you can browse operations and try them against sample data:
   claude mcp add --transport http tross-sandbox https://mcp.ontross.com/sandbox/mcp
   (use this editor's equivalent if you are not Claude Code). This is the sandbox, not my account: no sign-in, no real patient data.
2. Detect this project's language and package manager.
3. Install a client:
   - TypeScript / JavaScript: npm install tross
   - Python 3.10+: pip install tross
   - Other languages: generate a client from https://api.ontross.com/openapi-sdk.json with openapi-generator, or call the REST API directly:
     POST https://api.ontross.com/api/integrations/{provider}/{operation}
     header X-API-Key: <key>, JSON body {"auth_id": "...", "input": {...}}
4. Configuration: add TROSS_API_KEY to .env.example with a placeholder, confirm .env is gitignored, and read the key from the environment in code. Then ask me to create a key at https://app.ontross.com/api-keys and put it in .env myself, and wait for my confirmation.
5. Verify: ask me for an auth_id (listed at https://app.ontross.com/auth), write a small script that calls one read-only operation with it, run it, and show me the response status and the shape of the output, never the key. 401 means the key is wrong; 404 means the provider or operation name is wrong, check its docs page.
6. Finish with a short summary: what you installed, which files changed, and the one command I run to make my first call.

What the assistant does

  1. Connects its toolsAdds the Tross sandbox MCP server so it can browse every operation and try them against sample data.
  2. Detects your stackLooks at the project to pick the right client: the npm or PyPI package, or the REST API for any other language.
  3. Installs and configuresAdds the package, a TROSS_API_KEY placeholder in .env.example, and reads the key from the environment.
  4. Keeps your key out of the chatAsks you to create the key and put it in .env yourself. It never prints or reads the value.
  5. Verifies with a real callRuns one read-only operation with an auth_id you choose and shows you the response shape.

Prefer to do it yourself

TypeScript / JavaScript

npm install tross
import { Configuration, IntegrationsApi } from "tross";

const tross = new IntegrationsApi(
  new Configuration({ apiKey: process.env.TROSS_API_KEY }),
);
const run = await tross.ezdermListAssessmentOptions({
  authId: "<auth_id>",
  input: { patientId: "...", encounterId: "...", providerId: "..." },
});

Python 3.10+

pip install tross
import os
from tross import ApiClient, Configuration
from tross.api import IntegrationsApi

tross = IntegrationsApi(ApiClient(Configuration(
    api_key={"ApiKeyAuth": os.environ["TROSS_API_KEY"]},
)))
run = tross.ezderm_list_assessment_options({
    "auth_id": "<auth_id>",
    "input": {"patient_id": "...", "encounter_id": "...", "provider_id": "..."},
})

MCP for your editor

Lets the assistant browse operations and try them against sample data before writing code. This is the sandbox: no sign-in, no real patient data.

claude mcp add --transport http tross-sandbox https://mcp.ontross.com/sandbox/mcp
Other editors, and your own account

No library for your language?

Generate one from the OpenAPI spec, or call the REST API directly: every operation is a POST with an X-API-Key header and a { auth_id, input } body.

https://api.ontross.com/openapi-sdk.json
npx @openapitools/openapi-generator-cli generate \
  -i https://api.ontross.com/openapi-sdk.json -g java -o tross-java