Three new sandbox providers for safely running agent code in isolated cloud containers.
Improve tool-call accuracy, MCP fetch hooks receive RequestContext, and reliability updates surface streaming errors, preserve history in stateless tool runs, and clean up orphaned vector embeddings.
Token-aware workspace output, smarter truncation, LSP resolution, sandbox process control, end-to-end auth and RBAC, workflow path tracking, concurrent-safe snapshots, and harness.sendMessage() file support.
A new video course designed to get you from zero to deploying your first agent in under two hours.
Run test cases against agents and workflows, score the results, and track quality over time.
Supervisor pattern for multi-agent coordination, metadata-only vector queries, more flexible runEvals options, LSP diagnostics after workspace edits, and a new Blaxel sandbox provider.
An open-source TUI coding agent with observational memory that compresses context without losing it.
Background process management in workspaces and sandboxes, runtime tool configuration via Workspace.setToolsConfig(), and observational memory reliability and introspection improvements.
AST-based workspace edits, real-time streaming tool argument previews, built-in task tracking tools for Harness, and improved Observational Memory continuity
Define test cases, run experiments, and track response quality for every change you make.
A new reusable Harness orchestration layer for agent apps, versioned workspaces and skills, pluggable blob storage, security and discovery upgrades, plus improved tool I/O and streaming behavior, including better chunk handling.
Learn how Mastra is optimized for modern AI systems, and how to implement a similar approach.
