Anthropic Unveils Multi-Agent Orchestration Architecture, Letting Claude Write Its Own Code to Manage AI Agents
The orchestration layer is now generated by the model at runtime rather than hard-coded by developers, a shift that complicates static auditing of agent-initiated payment paths.
Anthropic has announced a new-generation multi-agent orchestration architecture, whose core change is that Claude no longer serves only as a model to be called, but generates program code itself to schedule and manage multiple AI agents. According to 36Kr, the architecture has the model write control logic at runtime, with these self-written programs determining the division of labor and invocation order among agents.
Previously, multi-agent systems typically relied on fixed orchestration frameworks pre-written by developers, with the number of agents and task paths determined before deployment. Anthropic's approach hands the orchestration layer to the model for dynamic generation, which it describes as able to adapt to changes in task structure.
The announcement did not disclose a specific release timeline, availability scope or pricing information. For agent payment and settlement scenarios, having the orchestration layer generated by the model itself means that the invocation path behind each agent-initiated call is harder to audit statically before deployment, and the points at which authorization intent and per-transaction limits are verified must be determined alongside the generated logic.
Source: https://news.google.com/rss/articles/CBMiU0FVX3lxTE5Mc3F2cFc2OHlDdkJaTjZRblFMcTJqeUpIT29BcXhXYVduUWctSE1ydXpha1pXU1kxZWNHcDVVRGpRZmUyZjRwSW5PUjdrOHZPczRR?oc=5
Provenance & status
- Byline
- OceanAlt Editorial
- First published
- 2026-10-20
- Last updated
- 2026-10-20
- Content type
- Newsflash
- Source material
- View original ↗
Related reading
Harbor Enters AI Agent Insurance With $10M Seed Round, Leaving Liability Attribution and Claims Pricing as the First Unsolved Problems
South Korea's President Confirms AI Role in Attacks on Seven Financial Firms: A Look at the Attack Chain
NVIDIA Open-Sources Agent Security Platform: Another Piece in the Agent Security Infrastructure, from Testing to Deployment
Paste a payee address before you pay and see whether it's on a sanctions list, through a mixer, or tagged for fraud.
This judgement can sit inside your own product
One line of code; it touches neither your CSS nor your JS. The same pre-settlement judgement can appear in your articles, on your wallet's confirmation screen, or as an endpoint your agent calls before it pays.