3 min readEnglish
This week in AI: self-hosted agents, faster product shipping, and free Claude for teachers
This week brings news that matters for managers steering AI adoption: a new self-hosted option for running AI agents securely inside company infrastructure, real-world examples of companies shipping faster with AI coding tools, and a free Claude offer for US teachers. We also flag a Claude Code update that is developer-only, but useful to know about.
Claude Platform
Claude Managed Agents can now run entirely on your own company servers
Anthropic launched a self-hosted version of Claude Managed Agents, meaning companies can now run the whole agent service, including where agents, sessions and history are stored, on their own infrastructure instead of Anthropic's. This matters for teams with strict data or compliance requirements, since it gives more control over where sensitive information lives while still using Anthropic's AI models by default. Access currently requires contacting your Anthropic account team.
OpenAI
Amplitude uses an AI agent to ship product changes in 1 to 2 days
Amplitude built a proactive product agent called Wave, using OpenAI models, that identifies what to build, helps teams ship changes with coding agents, and measures the impact afterward. For managers, this is a concrete example of how AI agents can shorten the time between an idea and a shipped product change, moving some product management tasks from weeks to days.
Holiday Extras turns customer knowledge into new apps using Codex
Travel company Holiday Extras is using OpenAI's Codex to build new features faster, reporting higher booking conversion and new growth opportunities. This shows how a non-technical business can use AI coding tools to translate customer expertise directly into working products, without needing a much larger engineering team.
Codex helps Gondola raise lead engineer output by 50% as revenue doubles
Gondola, a platform that helps creatives get credit and get hired, reports that Codex raised its lead engineer's output by 50% and that ChatGPT simplified daily work across the company, alongside doubled revenue. It is another real example of AI tools changing how much a small technical team can deliver, relevant for managers evaluating where to invest in AI tools.
Anthropic
Anthropic gives US teachers free access to premium Claude features
Verified US K-12 teachers can now get free access to premium Claude capabilities, a library of teaching skills, and curricula aligned to standards in all 50 states. This is a sector-specific offer, useful mainly for education teams or anyone managing partnerships with schools.
Claude Code
Claude Code update adds experimental Code Mode and usability improvements (for developers)
The latest Claude Code release adds an experimental opt-in Code Mode where Claude writes JavaScript to call its tools, plus smaller usability changes like clearer warnings and a progress indicator during long-running tasks. This update is aimed at developers configuring Claude Code directly, not something most non-technical team members will need to touch.
Model Context Protocol
MCP, the standard behind many AI agent integrations, moves to a fixed twice-a-year release schedule
The Model Context Protocol, a widely used standard that lets AI agents connect to tools and data, will now publish specification updates on a predictable schedule, generally the last Tuesday of March and September. This mostly concerns developers building on MCP, but it is useful context for managers: tools built on this standard should become more stable and predictable to plan around over time.
Summary prepared by Wisly AI from the official sources linked above. If something affects you, open the source before deciding.