I
Agentic Intelligence · Infomly
Jun 15, 2026
5:02 AM
Agentic AI

Mastra just wrapped OpenAI Agents SDK. Your agent harness just became optional.

Mastra (25.1K stars) shipped @mastra/openai on June 12.

One import. Your OpenAI Agents SDK agent runs through Mastra's generate() and stream() with full tracing, usage continuity, and tool-call telemetry.

No rewrite. No new abstractions.

The @mastra/openai package exports OpenAISDKAgent. Pass sdkOptions or wrap a pre-created agent. Mastra handles the rest: usage normalization, provider metadata, and tool-call spans for function, hosted, shell, and apply_patch tools.

Why this matters:

Mastra already had @mastra/claude for Claude Agent SDK and @mastra/cursor for Cursor SDK.

With @mastra/openai, they now wrap all three major coding agent SDKs under one interface.

You can run an OpenAI Codex agent, a Claude Code agent, and a Cursor agent in the same session. Same tracing. Same memory. Same observability.

The interoperability layer the agent ecosystem has been missing.

They also shipped:

- @mastra/railway for Railway Sandboxes (cloud execution)
- @mastra/voyageai for VoyageAI embeddings and reranking
- Agent-agnostic task tools (task_write, task_write, task_complete) that work on ANY agent, not just Mastra's Harness
- Agent-agnostic ask_user and submit_plan via native tool suspension

The breaking change: harness.respondToQuestion is gone. Everything runs through tool_suspended events now.

If you're building multi-agent systems in TypeScript, evaluate Mastra as your orchestration layer. The SDK adapter pattern means you don't have to pick one vendor.

npm install @mastra/openai @mastra/claude @mastra/cursor

Run all three. Compare output. Pick the one that works.
SOURCE: https://github.com/mastra-ai/mastra/releases/tag/%40mastra/core%401.42.0
VERIFIED: GitHub release page, Mastra PR #17525, Mastra docs
SIGNAL: Mastra is positioning as the universal TypeScript meta-harness. With SDK wrappers for OpenAI, Claude, and Cursor, they're solving the vendor lock-in problem that every multi-agent team hits at scale.
---
1 views

0 Comments

No comments yet. Be the first.