Migrate a Workflow to Claude Managed Agents Safely
A safe rollout playbook for moving an existing workflow onto self-hosted Claude managed agents: shadow mode, canary, human-in-the-loop, and instant rollback.
9 min read
Agentic AI & LLMs
Agentic AI & LLMs archive page 275 of 631
Agentic AI & LLMs
Agentic AI, LLM engineering, and the models behind modern automation — multi-agent systems, LLM evaluation and comparisons, RAG, fine-tuning, AI infrastructure, security, and production AI engineering.
From the blog
Latest articles
Evals for Claude Agents: Measure Quality, Gate Releases (Managed Agents Sandboxes Tunnels)
Build an eval loop for self-hosted Claude managed agents: trajectory and outcome metrics, LLM-as-judge graders, golden datasets, and CI release gates.
8 min read
Securing Claude Managed Agents: Sandboxing & Least Privilege
Harden self-hosted Claude managed agents with sandbox isolation, least-privilege MCP tools, server-side secrets, and layered prompt-injection defense.
9 min read
Cut Claude Agent Token Cost: Caching and Batching
Make self-hosted Claude managed agents cheap and fast with prompt caching, batched tool calls, model routing, and context hygiene. Concrete tuning steps.
9 min read
Debugging Claude Managed Agents: Loops, Bad Tool Calls (Managed Agents Sandboxes Tunnels)
Trace and fix the three failure modes of self-hosted Claude managed agents: runaway loops, wrong tool calls, and hallucinated arguments over MCP.
9 min read
Context Design for Claude Managed Agents: What to Include
What to put in a Claude managed agent's context and what to leave out: ordering, just-in-time retrieval, compression, and untrusted-content safety.
8 min read
Wiring MCP Servers into Claude Agents: Auth and Idempotency
Auth, schemas, error handling, and idempotency for wiring MCP servers into Claude managed agents so tool calls survive retries safely.
8 min read
Claude Managed Agent Patterns: Tools, Prompts, and Context
Reusable patterns for tools, prompts, and context in Claude managed agents: verb tools, discovery-first, layered prompts, and context budgeting.
8 min read
Build a Claude Managed Agent: Sandbox and MCP Tunnel Setup
A hands-on walkthrough to stand up a self-hosted sandbox, run an MCP server, open an MCP tunnel, and drive a Claude managed agent.
8 min read