Launch HN: Inkeep (YC W23) – Agent builder that works both visually and in code (github.com)

Inkeep (YC W23) launched an open-source, source-available framework for building and shipping AI Agents that blends a no-code visual builder with a TypeScript SDK and two-way sync. It targets both chat assistants (customer support, in-app copilots, internal team copilots) and workflow automation (knowledge base updates, CRM edits, ticket triage). The platform is modular: agents-manage-api/UI for configuration and visual editing, agents-sdk for declarative TypeScript agent definitions, agents-run-api as the runtime that exposes Agents as APIs and maintains conversation state, a CLI for push/pull sync, a UI component library for chat embeds, and observability via a Traces UI + OpenTelemetry. Deployments support Vercel and Docker, and triggering/integration options include MCP, agent-to-agent (A2A), and Vercel SDK APIs. Technically, Inkeep emphasizes multi-agent architectures, credentials management, and interoperability: the visual builder and code-first SDK are fully interoperable so engineers and non-engineers can collaborate on the same agents. Under the hood it leverages the Vercel AI SDK for LLM access and emits streams compatible with Vercel useChat/AI Elements, while remaining LLM-agnostic and self-hostable. The project is released under Elastic License 2.0 with supplemental terms (fair-code, source-available) — important context for teams weighing open extensibility against usage restrictions.