Unified model + tool interface

The TypeScript toolkit we reach for to wire up streaming, tool calls, and model swaps across providers.

Durable agent execution

Long-running, resumable workflows for building agents that survive restarts, retries, and human-in-the-loop steps.

Autonomous loops with Claude

Anthropic’s agent runtime — tool use, memory, and long-context reasoning wrapped in a production-ready SDK.

Hosted agent infrastructure

Managed deployment for Claude-powered agents so we can ship without running the runtime ourselves.

Model Context Protocol

The open protocol for connecting models to tools and data sources. We build MCP servers and clients to extend agent capabilities.

Scheduled and event-driven jobs

Cron, triggers, and event pipelines that let agents and products act on their own cadence — without a human in the loop.

Agent tooling framework

Open-source tooling we use and contribute to for building, evaluating, and deploying agent-native software.