Runtime for defining and executing portable AI workers. Multi-provider LLM support (Anthropic, OpenAI, Gemini, and OpenAI-compatible vendors), pluggable persistence and tracing adapters, an MCP tool client, streaming, abort, and pause/resume.

Type-safe error management using generators. Inspired by [EffectTS](https://effect.website/) and [typescript-result](https://github.com/everweij/typescript-result)

OpenTelemetry-compatible tracing for AgentMug worker execution. Emits LLM-call and transcription spans/events into your existing OTel SDK pipeline (Honeycomb, Datadog, Jaeger, Tempo, Sentry).

Framework-agnostic tool-call reliability boundary for LLM agents and gateways.

Security utilities — validation, audit logging, sanitisation, env guards

GraphQL authorization directive

React reusable components for VODO frontend developers

A random username generator with customization options.

Drop breadcrumbs from your AI conversations to Slack and beyond

Download a gosd-built SD-card image, verify it, and splice per-deployment config into its reserved placeholder files as it streams to disk.

A plug-and-play AI Auditing & Hallucination Detection SDK. Intercepts LLM inputs/outputs, verifies factual grounding against your knowledge base, and blocks or flags unverified responses.

MCP server exposing 2slides.com tools (slides generate, jobs get, themes search)

The Simple Object Validator.

SDK for Trainkore AI platform

Converts a graphql schema to a JSON object

A defense library for LLM-based apps to detect and block prompt injection attacks

SDK ELYSÉA — cerveau éthique cross-app. Guardian SDK, pipeline canon 9 étapes, mémoire souveraine, identité universelle. TypeScript pur, zero browser deps.

Shared East type definitions for e3-aws authorization

No description provided.

Sanitize AI prompts. Scan, armor, and parameterize untrusted input before it touches your LLM — like prepared SQL statements, but for AI.