Agentuity

The Agentuity squad checking in from React Miami Day 1! Come swing by our booth and say hello!

The Agentuity squad checking in from React Miami Day 1! Come swing by our booth and say hello!

Coding agents are standard now. The next question is what teams do with them. At 11:20 AM ET, our co-founder Rick Blalock is speaking at AI Engineer: Miami, covering "how coding agents ate the world", and about code as the execution harness for agents. Tune into the livestream to watch from anywhere: https://lnkd.in/gUgC-eFW

SDK v2 is here! Dev mode was slower than we wanted, so we dug in and rebuilt the runtime from the ground up. What changed: • Removed codegen and file-based routing in favor of explicit declarations • Standard Hono + Vite, so no proprietary wrappers • Dev mode is faster than ever, without the bundling overhead If you're already on v1, migrate with: bunx @agentuity/migrate Read more about the change: https://lnkd.in/g_mdwHhS

Agentuity services are now available as their own standalone packages, with the same APIs you already know from agents and routes! Use them from any Node.js or Bun app. For example, spin up an Agentuity sandbox from a Bun script, write code to it, and run it. Available as standalone packages: • Sandboxes for isolated code execution • Vector search with auto-embeddings • Queues, email, webhooks, and more Docs: https://lnkd.in/geuGs6vU

We built OAuth 2.0 and OIDC into the Agentuity platform, so your project can be the identity provider. No external auth service needed. Register an app and you get: • Redirect URIs and granular scope permissions • Control over which org members have access • Full authorization code flow: consent, token exchange, and user info • Scoped access to Agentuity APIs on behalf of users Available from the dashboard or the CLI. Try it today: https://lnkd.in/g3PfFFbN

We've got new examples walking you through how to use Mastra on Agentuity! Six agents, from conversational memory to multi-agent networks with human-in-the-loop approvals. Each one follows the same pattern: bring your own framework, plug into the services you need — thread state, agent-to-agent networking, observability, and more. Guide: https://lnkd.in/gkTk__Zj Code: https://lnkd.in/gREzqJ4G

We built the SDK Explorer as the landing page for our documentation, and we designed it around one idea: learn by doing. Each page pairs an interactive demo with full reference code, and you can run that code in a real Agentuity sandbox. You get to try out the same services your agents use when deployed 🤖 The Explorer covers demos across the SDK, including: • Agent context, state management, and background tasks • Storage: KV, vector search, and object storage • Streaming: SSE, WebSockets, and durable streams • Built-in evals with scoring and verification Give it a try: https://agentuity.dev Full video: https://lnkd.in/gjfnyaQu Blog post: https://lnkd.in/gfFF4S36

Agentuity sandboxes can take one prompt, generate TypeScript and Python code, and run both in isolated environments. Plus, you get end-to-end type safety across the agent, API, and frontend. Our built-in evals suite can then validate the sandbox output with an LLM-as-a-judge pattern. See the full demo: https://lnkd.in/gqTw65Ny Check out the code: https://lnkd.in/gXz7PKGp