Packages

Convention and validation for AI agent runbooks

A Rust implementation of the Model Context Protocol (MCP) for AI tool integration

Internal reusable Agent Runtime scaffolding, validation, and packaging services

Internal production composition root and lifecycle bootstrap for Agent Runtime

Unified CLI for serving, creating, inspecting, testing, and building Agent Runtime definitions

Reference production Code Agent for Agent Runtime

Validated deployment configuration for Agent Runtime

Internal Runtime-owned control tools

Internal immutable Agent definition builder

Internal safe Agent definition loader and catalog

Internal durable execution coordinator for Agent Runtime

Internal execution lifecycle contract for Agent Runtime

Internal extension contract backing the Agent Runtime SDK

Internal HTTP API contract for Agent Runtime

Internal HTTP and SSE transport for Agent Runtime

Transport-neutral contracts for Agent Infra runtime identity APIs

Internal Agent Infra SDK adapter for Agent Runtime ports

Internal model, context, and Tool execution Kernel for Agent Runtime

Internal Kernel contract for Agent Runtime

Internal in-memory execution journal for Agent Runtime

Internal safe Agent package loader and catalog

Internal provider-neutral ports for Agent Runtime

Stable SDK for building Agent Runtime agents, tools, context hooks, guards, and observers

Deterministic test fixtures and assertions for Agent Runtime extensions

Internal value types for Agent Runtime

SPL (Safe Policy Lisp) evaluator for Agent-Safe capability tokens. 150 lines, zero deps core, microseconds.

A sandboxed execution environment for AI agents via WASM

Scaffold the agent workflow into a project.

Scratchpad for intermediate reasoning steps in LLM agents

Canonical byte-deterministic transcript format for AI-agent conversations (Rust port of @p-vbordei/agent-scroll)

Rust Agent SDK for building LLM agents

Developer-experience CLI for the Agent SDK (e.g. installs the local Langfuse + OTel collector stack)

Product-neutral primitive kernel and contracts for a Rust-first Agent SDK.

Optional evaluation framework primitives layered over agent-sdk-core traces and evidence.

Shared contract types for the Agent SDK (IDs, events, LLM messages, turn outcomes)

Ergonomic derive/declarative macros for the Agent SDK (#[derive(Tool)], #[derive(TypedTool)], tool!, #[derive(ToolName)])

OpenTelemetry bootstrap helpers for the Agent SDK (OTLP exporter wiring + propagators)

Optional provider adapter helpers layered over agent-sdk-core ProviderAdapter contracts.

LLM provider trait, streaming primitives, and first-party provider implementations for the Agent SDK

Pure-Rust agent SDK for production tool-calling loops, multi-provider, type-safe, and ergonomic

Filesystem-backed durable store adapters for the Agent SDK.

Scripted Postgres-style durable store adapters for the Agent SDK.

SQLite-backed durable store adapters for the Agent SDK.

Supabase REST-backed durable store adapters for the Agent SDK.

Optional tool-pack helpers layered over agent-sdk-core primitive contracts.

Tool traits, registry, hooks, and store contracts for the Agent SDK

Unified multi-provider search CLI for AI agents — 13 providers, 13 modes, email verification, one binary

App-scoped Wayland capture and input for Linux automation agents

Transport-neutral secret reference contract shared across Agent Infra domains

Ahp public package coordinate scaffold