Claude Mythos and Conway: Anthropic's New Model and Always-On Agent Explained
April 8, 2026 · 7 min read · Happycapy Guide
Anthropic is testing two major releases in April 2026: Claude Mythos, its next flagship model, and Conway, an always-on AI agent designed to run continuously and complete multi-step tasks autonomously. Both are in internal testing. Here is what has been reported and what it means for AI power users.
Anthropic had a remarkable week. The company publicly refused to deploy Claude on US Department of Defense classified networks — a decision that drove Claude to the number-one spot on the US App Store as #QuitGPT users sought alternatives to ChatGPT. Simultaneously, internal reports surfaced on two major forthcoming releases: a new flagship model codenamed Claude Mythos and an always-on autonomous agent called Conway. Here is everything confirmed so far.
What Is Claude Mythos?
Claude Mythos is Anthropic's next flagship model, reported to be in internal testing and expected to launch in April 2026. It is the successor to the Claude 3.x family (Haiku, Sonnet, Opus). While Anthropic has not published a spec sheet, sources describe Mythos as targeting three specific capability areas that the current Claude lineup underperforms on compared to GPT-5.4:
- Long-horizon task planning — multi-step reasoning across very long contexts without losing coherence
- Agentic tool use — improved accuracy when chaining tool calls, web searches, and code execution in a single session
- Instruction following under adversarial conditions — maintaining safety and accuracy when users attempt to manipulate model behavior
The timing is strategic. If Mythos launches in April 2026 while Claude is benefiting from the #QuitGPT migration surge, Anthropic will be introducing its most capable model to its largest-ever user base. That is a favorable launch window by any measure.
“Mythos” appears to be an internal codename, consistent with Anthropic's history of using thematic names during development (prior models were codenamed with similar mythological or philosophical references). The public release name may differ.
What Is Conway?
Conway is Anthropic's always-on AI agent — and the most significant architectural departure from the standard Claude chat interface. Unlike Claude's current model, where you open a conversation, ask a question, and receive an answer, Conway is designed to operate continuously: running in the background, monitoring conditions, and executing multi-step tasks autonomously with minimal user input.
Runs continuously in the background — not just when you open a chat window. Monitors inputs and acts on triggers.
Completes complex, chained tasks without step-by-step human guidance. Sets sub-goals and executes toward them.
Built to surface only when it needs input — not to demand constant confirmation. Designed for low-touch workflows.
Expected to run on Anthropic's next-generation model for advanced reasoning and long-horizon task planning.
Conway is Anthropic's answer to a growing category of persistent AI agents: Cognition's Devin (autonomous coding), OpenAI's Operator (browser-based task completion), and agentic platforms like Happycapy. The “always-on” framing is key — it is not about a smarter chatbot, it is about an AI that does not require you to be present for every step of a workflow.
Always-On AI Agents Are Available Today on Happycapy →Timeline: What We Know and When
Internal testing of Conway confirmed by sources familiar with the project. Always-on agent architecture described in technical briefs.
Reports of Claude Mythos testing surface alongside Conway. Anthropic refusal of DoD deal drives Claude to #1 on US App Store — positioning Mythos launch in favorable market conditions.
Claude Mythos public release expected. No official date confirmed by Anthropic at time of writing.
Conway always-on agent expected to move from internal testing to limited beta, then general availability.
Claude Mythos vs. Conway: What's the Difference?
It is important to distinguish between these two releases. Claude Mythos is a model— a more capable version of Claude's intelligence layer. Conway is an agent architecture — a framework for how that intelligence is deployed persistently over time. Mythos makes Claude smarter. Conway makes Claude always-on. They are complementary, not competing.
The likely path is that Conway runs on Mythos once both are available — combining Anthropic's most capable model with an always-on execution layer. If that combination delivers on its description, it would represent a meaningful competitive upgrade against GPT-5.4 / Operator from OpenAI.
How This Compares to What Happycapy Offers Today
| Capability | Claude (current) | Conway (testing) | Happycapy (now) |
|---|---|---|---|
| Persistent / always-on operation | No — session only | Yes — in testing | Yes — production |
| Multi-step autonomous tasks | Limited — within session | Yes — designed for it | Yes — 150+ skills |
| Email automation | No | TBD | Yes — Capymail |
| Mac desktop control | No | TBD | Yes — Mac Bridge |
| Async task delivery | No | TBD | Yes — email on completion |
| Publicly available | Yes | No — internal testing | Yes |
| Price | $20/mo (Pro) | TBD | $17/mo (Pro) |
| Model | Claude 3.x | Claude Mythos (expected) | Claude + multi-model |
The honest comparison: Conway is promising but not yet available. Happycapy already ships what Conway is designed to do — persistent agentic workflows, async task delivery, and production-grade automation skills — available today on the Pro plan at $17/month. When Conway launches, it will bring Anthropic's infrastructure-level agent natively into Claude Pro and Max. Until then, Happycapy is the most capable production option for users who want always-on AI workflows powered by Claude.
Try Happycapy — Always-On AI Agents, Available Now →Frequently Asked Questions
What is Claude Mythos?
Claude Mythos is Anthropic's next flagship model, reportedly in internal testing and expected to release in April 2026. It targets long-horizon reasoning, agentic tool use, and improved instruction following — succeeding the Claude 3.x family.
What is Conway by Anthropic?
Conway is Anthropic's always-on AI agent, currently in testing. It is designed to operate continuously, monitor inputs, and complete multi-step tasks autonomously with minimal user interaction — similar to OpenAI's Operator but built on Claude.
How does Conway compare to Happycapy?
Conway is a research-stage always-on agent from Anthropic, not yet publicly available. Happycapy already ships a production-grade agentic platform with 150+ skills, async task delivery via Capymail, persistent memory, and Mac Bridge for desktop control — available today for $17/month on the Pro plan.
When will Claude Mythos be released?
Claude Mythos is reportedly in internal testing and expected to launch in April 2026. Anthropic has not made an official public announcement with a confirmed release date as of April 8, 2026.
What is an always-on AI agent?
An always-on AI agent runs persistently rather than responding only when directly prompted. It monitors conditions, triggers on events, completes multi-step workflows autonomously, and delivers outputs asynchronously — without the user needing to be present for every step. Happycapy's Pro plan includes this capability today.
Crescendo AI — “Latest AI News and AI Breakthroughs that Matter Most: 2026” (April 4, 2026)
Radical Data Science — “AI News Briefs Bulletin Board for April 2026” (April 3, 2026)
LLM Stats — “AI Updates Today (April 2026)” (April 8, 2026)
Mean CEO Blog — “New AI Model Releases News — April 2026 (Startup Edition)” (April 1, 2026)
PCMag — “Anthropic: You Can't Use OpenClaw With Claude Without Paying Extra” (April 7, 2026)