Strands Agents 1.58.1 fixes MCP task compatibility and nested agent interrupts
The Apache-2.0 agent SDK’s October 6 patch restores legacy MCP task execution for SDK 2.x clients, fixes nested agent-as-tool interrupt resumption, and preserves reported cache-token usage.
2 min read
Strands Agents released Python SDK 1.58.1 on October 6, 2026, with a small set of fixes aimed at agent runtimes that depend on MCP, nested agents and accurate usage reporting.
The patch restores legacy task execution when using the SDK 2.x MCP client. It also fixes propagation and resumption of interrupts raised inside an agent invoked as a tool. That matters for human-in-the-loop workflows: a parent agent should be able to pause for approval or input inside a nested agent, then continue without losing the interruption state.
A separate model-accounting fix preserves cache-token usage reported by providers. Applications that use those figures for cost attribution or observability should see more faithful totals after upgrading. The release also changes the TypeScript harness so @strands-agents/sdk is a peer dependency. Although the tag is named python/v1.58.1, the monorepo’s automatically generated notes include these cross-package TypeScript changes, so teams should check the package they actually deploy.
Strands runs the agent loop inside the developer’s process and supports MCP, tools, memory, sessions, guardrails and multiple model providers. The repository is licensed under Apache 2.0. Its Python SDK requires Python 3.10 or newer; the TypeScript SDK requires Node.js 22 or newer.
This is a maintenance release rather than a new architecture or capability launch. Upgrading is most relevant for teams already using MCP 2.x compatibility, nested agent-as-tool patterns, or provider cache-token metrics.
Sources: Strands Agents 1.58.1 release notes — October 6, 2026 · Strands Agents repository, installation requirements and Apache 2.0 license · Unsplash image license
Conversation
0 approved commentsNo comments yet. Start a thoughtful conversation.
Join the conversation.
Sign in with ChatGPT to leave a comment or rate this article.