Product Updates
Built an Agent? Let Your Team Use It.
Workspaces gives your team a home for the agents, threads, skills, and memories it shares.
Product Updates
Workspaces gives your team a home for the agents, threads, skills, and memories it shares.
product-updates
Your agent can carry work from research to action, bring context into new conversations, and show you what it plans to do, remember, and learn.
New this week: Four new models ready to take on any job, agents can search across team documents, tables, and memories for shared context. Plus, safer agent updates, and an improved experience with Slack. * Gemini 3.7 Flash fits document-heavy operations. Think: reviewing ten vendor proposals, comparing them against
Product Updates
New this week: stronger models for the agents that coordinate multi-step jobs and pass work to other agents. Plus, clearer handoffs between agents, and alerts when an agent needs you.
Hyperagent has everything your OpenClaw setup was trying to do, without the burden of running the infrastructure underneath it.
Product Updates
New this week: a practical guide to choosing the right model for your agent's work. Try the newest models available: Gemini 3.6 Flash, and Inkling. Plus, switch models directly from thread headers, smarter thread tabs, shared billing, and complete search in Hyperagent via Cmd + K. Which model
The best model for a job is almost never the best model on the leaderboard. Fourteen models, forty-two runs, on the real knowledge work our customers hand to agents every day.
Last updated: May 26, 2026 at 16:00 PT. We will continue to update this page as we learn more, so please check back regularly. On May 21, 2026, Composio, a platform we use to facilitate integrations between Hyperagent and various third-party data sources, disclosed a security incident in
11 agents manage Deskimo's BD pipeline. The agent that runs today is meaningfully different from the one they shipped. Same prompt, same integrations, but three weeks of editorial judgment get built into every draft using a learning loop powered by Hyperagent's memories.