Product updates
Here we bring together changes to agents, Brains, tool access, observability and evaluations. Each update explains what changed and why it matters to operational teams.
- New connectors and MCP capabilities
- Improvements to logging, evaluations and regression tests
- Practical release notes for administrators
Engineering notes
We write about architectural choices such as secure tool execution, human-in-the-loop approval, OAuth scopes and semantic search across company knowledge.
- How to limit agents to the right tools
- Why evaluations for each workflow matter more than general benchmarks
- How context from Brains is added to prompts in a controlled way
Implementation stories
From the first pilot to rollout across multiple workflows: we share recurring patterns in support, sales, back-office work and internal helpdesks.
Best practices
Short guides to prompting, escalation rules, data sources, monitoring and designing workflows that agents can carry out reliably.