Build AI cost evidence into every workflow
Integrate trAIce through maintained TypeScript and Python SDKs, the official CLI, a versioned REST API, or a remote MCP server. The hosted service receives usage evidence after model calls, so it does not need to sit in your request path.
Choose an integration surface
SDKs and CLI
Instrument TypeScript or Python applications after each model call. The official npm package also publishes the traice CLI for authentication and workspace-scoped cost questions.
REST API
Send usage evidence, query AI economics, and discover stable integration metadata through a versioned JSON API with typed errors and explicit rate-limit conventions.
MCP and OAuth
Connect an agent to the remote MCP server with OAuth discovery, least-privilege read and confirmed-write scopes, and machine-readable protected-resource metadata.
Self-serve onboarding
Start with a free workspace and no credit card. Generate a scoped API key in the application, install an SDK, and send a first event. Use the public status endpoint and OpenAPI document without authentication to inspect the contract before creating an account.
Create a free workspace, review the current free tier, then create a key from the self-serve API Keys page in the application. No sales conversation is required.
For agent integrations, start with the agent guide or the MCP server card. For REST integrations, review the versioning and deprecation policy before relying on a production operation.