Claude Fable 5, Managed Agents, Claude Code: All Anthropic News from June 2026
Anthropic dropped a massive wave of updates between June 8-10, 2026. From Claude Fable 5 (a Mythos-class model for the general public) to Managed Agents that schedule themselves and authenticate securely, plus a native Swift SDK for Apple developers—Claude’s ecosystem has never seen a week this packed.
Let’s break it all down.
---
🔥 Claude Fable 5: Mythos-class Goes Public
On June 9, Anthropic launched Claude Fable 5, the first Mythos-class model available to the general public. Until now, this tier was restricted to specific partners (cyberdefense, critical infrastructure) through Project Glasswing.
What’s New
| Feature | Claude Opus 4.8 | Claude Fable 5 |
|---|---|---|
| Tier | Standard | Mythos-class |
| Reasoning | Very good | Excellent (multi-step, complex code) |
| Safety routing | No | Yes — sensitive prompts routed to Opus 4.8 |
| Pricing | Included in subscriptions | $10/M input / $50/M output (after June 22) |
Safety Routing: The Key Innovation
Instead of restricting the model, Anthropic implemented intelligent routing: when a query involves sensitive domains (cyber, biology, chemistry, model distillation), it’s automatically redirected to Claude Opus 4.8. This lets Fable 5 stay as unrestricted as possible where it excels, while keeping a safety net.
Availability & Pricing
• Until June 22 : included for free in Pro, Max, Team, and seat-based Enterprise plans
• From June 23 : moves to usage credits ($10/M input tokens, $50/M output)
• API & AWS : available now on Claude API and Amazon Bedrock
Tip: If you’re a Claude subscriber, use these 12 days to test Fable 5 on your complex use cases. After June 22, the bill can add up quickly.
Claude Mythos 5: The Unrestricted Version
For cyberdefenders and critical infrastructure providers, Anthropic also released Claude Mythos 5 — the same model with cyber safeguards lifted. Limited to Glasswing partners (and soon a broader trusted access program).
---
⚙️ Claude Managed Agents: Agents That Work Solo (Even at Night)
On June 9, Anthropic unveiled two major features for Managed Agents, now in public beta.
1. Cron Scheduling Built-In
Agents can now run on a schedule without you needing to host a scheduler yourself.
Typical use cases: - 🗓️ Nightly data sync - 🔒 Weekly compliance scan - 📊 Daily automated report - 🚨 Hourly uptime check
How it works: Define a cron schedule, and the Managed Agent starts a new session at each trigger. Pause, resume, or archive deployments anytime.
2. Vaults for CLIs and Authenticated Tools
Agents can now use environment variables stored in vaults to authenticate with CLIs and external tools.
Enhanced security: - 🔑 The agent never sees the key — a placeholder runs in the sandbox - 🛡️ The real key is attached at the network boundary, only on approved domains - 🔄 Update a key in the vault and running sessions pick it up on their next call
Compatible tools: Browserbase (web navigation), KERNEL (databases), Notion, Ramp, Sentry, and any CLI that sends keys via HTTP requests.
---
🍎 Claude on Apple: Foundation Models SDK
On June 8, Anthropic released a Swift package letting Apple developers use Claude directly from Apple’s Foundation Models framework.
3 Lines of Code
import Claude
let result = try await Claude.analyze("Summarize this contract in 3 points")
// → Returns a typed Swift value directly
What It Means
• iOS 27, macOS 27, visionOS 27, watchOS 27 supported
• Quick tasks (summarization, extraction) stay on Apple’s on-device model
• Complex tasks (multi-step reasoning, code generation, web search) are handed off to Claude
• Streaming and tool calls are handled natively by the package
---
🛠️ Claude Code v2.1.163
The Claude CLI was updated with several useful features:
• /plugin list — List installed plugins with --enabled/--disabled filters
• Improved hooks — Stop and SubagentStop hooks can return additional context without being flagged as errors
• Skills: \$ escaping — Include literal $ in command bodies
• Fix: claude -p — No longer hangs forever when a backgrounded command never exits
• Fix: Background sessions in claude agents no longer lose running tasks when reattached
---
Summary
| Feature | Date | Impact |
|---|---|---|
| 🔥 Claude Fable 5 | June 9 | Public access to Mythos-class model |
| ⚙️ Managed Agents (scheduling + vaults) | June 9 | Fully autonomous agents |
| 🍎 Apple Foundation Models SDK | June 8 | Native Claude on Apple devices |
| 🛠️ Claude Code v2.1.163 | June 4 | Plugins, hooks, bug fixes |
Bottom line: June 2026 is the month Anthropic stepped on the gas. A flagship model open to the public, agents that work on their own, and native Apple integration—Claude’s lineup has never been this complete.
---
Want to leverage these advances?
At Izri.Online, we track these developments daily. Whether you want to integrate Claude into your stack, automate processes with agents, or simply choose the right model for your project, we can help.
👉 Book your free audit — We analyze your digital presence and propose a personalized action plan. 2 Humans + 10 AI Agents at your service.
---
Created by Izri.Online — 2 Humans + 10 AI Agents Morocco · France · Africa · Belgium · Switzerland
