UponAI developer tools

@upon-ai/sdk and @upon-ai/mcp-server, owned by UponAI and licensed under MIT.

npm install @upon-ai/sdk@0.1.7

SDK guide · MCP setup · Interactive API

This independent release repository consumes the reviewed customer OpenAPI contract. Internal application docs remain in the application repository. No administrative APIs or privileged impersonation headers are included.

Build and verify

Requires Node >=22.14. Use the pinned lockfile:

npm ci --ignore-scripts
npm run check
npm run pack:release

Generated sources are reproducible from contract/openapi.json and its explicit policy. The SDK has 133 typed operations. MCP exposes 36 curated read operations plus workspace discovery, and twelve separately enabled write operations for calls, simulation sessions, improvement runs, schedules, agent webhook settings and agent environment tag pointers. Webhook settings use the platform proxy and saved customer selections.

• SDK usage

• MCP configuration

• Release and npm setup

• Security boundary

The MCP is a fresh implementation sharing this SDK. The earlier prototype was reviewed as a reference; its raw request escape hatch and dashboard phone tools are not carried into these packages. No third-party prototype source is copied.

Tests use isolated mock API responses and a real MCP protocol client. Live customer-workspace integration is a separate release gate; a passing local test suite is not evidence that all API operations work in production.

Recurring voice improvement policies are available through getVoiceSimulationSchedule, saveVoiceSimulationSchedule and pauseVoiceSimulationSchedule. A policy requires explicit review, per-cycle budgets and a bounded number of cycles before reapproval. MCP exposes schedule summaries and the optional pauseVoiceSimulationSchedule write capability; automatic recurring approval is deliberately excluded from the MCP catalog.